feat(subscribe): improve search scheduling feedback

app/api/dependencies/subscription.py

@@ -42,7 +42,10 @@ from app.application.subscription.mutation import (
     SubscriptionMutationService,
 )
 from app.application.subscription.query import SubscriptionQueryService
-from app.application.subscription.search import SearchSubscriptionsCommand
+from app.application.subscription.search import (
+    SearchSubscriptionsCommand,
+    SubscriptionSearchSubmission,
+)
 from app.application.subscription.status import (
     SubscriptionExecutionReadRepository,
     SubscriptionExecutionStatusService,
@@ -120,24 +123,14 @@ def get_delete_subscriptions_by_identity_command(
     )
 
 
-def _start_subscription_search_batch(
-    subscribe_ids: tuple[int, ...] | None,
-    state: str | None,
+def _resume_submitted_subscription_search(
+    subscribe_ids: tuple[int, ...],
 ) -> None:
-    """把一个请求的搜索目标作为同一调度任务提交。"""
-    if subscribe_ids is None:
+    """唤醒搜索队列，并优先处理本次手工选择的订阅。"""
     start_scheduler_job(
-            "subscribe_search",
-            sid=None,
-            state=state,
-            manual=True,
-        )
-        return
-    start_scheduler_job(
-        "subscribe_search",
-        sids=subscribe_ids,
-        state=None,
-        manual=True,
+        "subscribe_search_queue",
+        limit=max(1, len(subscribe_ids)),
+        manual_sids=subscribe_ids,
     )
 
 
@@ -146,23 +139,40 @@ def get_search_subscriptions_command(
     db: AsyncSession = Depends(get_async_session),
     runtime: HostRuntime = Depends(get_host_runtime),
 ) -> SearchSubscriptionsCommand:
-    """组装手工订阅搜索用例，并把调度延迟到响应后的后台任务。"""
+    """组装手工搜索用例，请求内只入队，实际搜索交给后台继续。"""
 
-    def schedule_search(
-        subscribe_ids: tuple[int, ...] | None,
-        state: str | None,
-    ) -> None:
-        """把当前用户的搜索目标提交为一个顺序后台批次。"""
-        resolve_background_task_registry(task_registry).create_sync(
-            _start_subscription_search_batch,
+    registry = resolve_background_task_registry(task_registry)
+    search_repository = get_subscription_search_repository(runtime)
+
+    async def submit_search(
+        subscribe_ids: tuple[int, ...],
+        single: bool,
+    ) -> SubscriptionSearchSubmission:
+        """在线程 owner 中完成轻量入队，并返回前端可立即跟踪的批次。"""
+        enqueue_task = registry.create_sync(
+            search_repository.enqueue,
+            subscription_ids=subscribe_ids,
+            source="manual",
+            priority=120 if single else 100,
+            owner="api.subscribe.search.enqueue",
+        )
+        enqueued = await enqueue_task
+        registry.create_sync(
+            _resume_submitted_subscription_search,
             subscribe_ids,
-            state,
-            owner="api.subscribe.search",
+            owner="api.subscribe.search.run",
+        )
+        return SubscriptionSearchSubmission(
+            batch_ids=enqueued.active_batch_ids,
+            target_count=len(subscribe_ids),
+            queued_count=enqueued.created_count,
+            ongoing_count=enqueued.coalesced_count,
+            single=single,
         )
 
     return SearchSubscriptionsCommand(
         repository=runtime.subscription.repository(db),
-        schedule_search=schedule_search,
+        submit_search=submit_search,
     )
 
 

app/api/endpoints/subexecution.py

@@ -32,7 +32,7 @@ async def _accessible_subscription_ids(
 
 @router.get(  # type: ignore[misc]
     "/execution/batches",
-    summary="查询订阅搜索批次状态",
+    summary="查看订阅搜索进度",
     response_model=List[_SchemaSubscriptionBatchStatus],
 )
 async def list_subscription_execution_batches(
@@ -52,7 +52,7 @@ async def list_subscription_execution_batches(
 
 @router.get(  # type: ignore[misc]
     "/execution/batches/{batch_id}",
-    summary="查询订阅搜索批次",
+    summary="查看一次订阅搜索",
     response_model=_SchemaSubscriptionBatchStatus,
 )
 async def get_subscription_execution_batch(
@@ -68,13 +68,13 @@ async def get_subscription_execution_batch(
         accessible_subscription_ids=accessible_ids,
     )
     if batch is None:
-        raise HTTPException(status_code=404, detail="订阅搜索批次不存在")
+        raise HTTPException(status_code=404, detail="没有找到这次搜索，请刷新后重试")
     return _SchemaSubscriptionBatchStatus.model_validate(batch)
 
 
 @router.put(  # type: ignore[misc]
     "/execution/batches/{batch_id}/cancel",
-    summary="取消订阅搜索批次",
+    summary="停止一次订阅搜索",
     response_model=_SchemaResponse[None],
 )
 async def cancel_subscription_execution_batch(
@@ -90,9 +90,9 @@ async def cancel_subscription_execution_batch(
         accessible_subscription_ids=accessible_ids,
     )
     if batch is None:
-        return _SchemaResponse(success=False, message="订阅搜索批次不存在")
+        return _SchemaResponse(success=False, message="没有找到这次搜索，请刷新后重试")
     cancelled = await status_service.request_cancel(batch_id)
     return _SchemaResponse(
         success=bool(cancelled),
-        message="" if cancelled else "订阅搜索批次已结束或无法取消",
+        message="" if cancelled else "这次搜索已经结束，暂时无法停止",
     )

app/api/endpoints/submaintenance.py

@@ -22,12 +22,50 @@ from app.application.subscription.mutation import (
 from app.application.subscription.search import (
     SearchSubscriptionsCommand,
     SubscribeSearchActor,
+    SubscriptionSearchSubmission,
 )
 from app.schemas.response import Response
+from app.schemas.subscribe import (
+    SubscriptionSearchSubmission as SubscriptionSearchSubmissionSchema,
+)
 
 router = ResponseAPIRouter()
 
 
+def _search_submission_message(submission: SubscriptionSearchSubmission) -> str:
+    """把搜索安排结果转换为用户能直接理解的提示。"""
+    if submission.target_count == 0:
+        return "没有需要搜索的订阅"
+    if submission.queued_count == 0:
+        return (
+            "这个订阅已经在搜索中，请稍候"
+            if submission.single
+            else f"{submission.ongoing_count} 个订阅已经在搜索中，无需重复提交"
+        )
+    if submission.ongoing_count:
+        return (
+            f"已安排 {submission.queued_count} 个订阅搜索，"
+            f"另有 {submission.ongoing_count} 个正在处理中"
+        )
+    if submission.single:
+        return "已安排搜索，很快开始"
+    return f"已安排 {submission.queued_count} 个订阅搜索，系统会依次处理"
+
+
+def _search_submission_schema(
+    submission: SubscriptionSearchSubmission,
+) -> SubscriptionSearchSubmissionSchema:
+    """构造稳定的手工搜索响应数据。"""
+    return SubscriptionSearchSubmissionSchema(
+        batch_id=submission.batch_id,
+        batch_ids=list(submission.batch_ids),
+        target_count=submission.target_count,
+        queued_count=submission.queued_count,
+        ongoing_count=submission.ongoing_count,
+        single=submission.single,
+    )
+
+
 @router.get(  # type: ignore[misc]
     "/refresh",
     summary="刷新订阅（兼容入口）",
@@ -97,38 +135,46 @@ def check_subscribes(
 @router.get(  # type: ignore[misc]
     "/search",
     summary="搜索所有订阅（兼容入口）",
-    response_model=Response[None],
+    response_model=Response[SubscriptionSearchSubmissionSchema],
     include_in_schema=False,
     deprecated=True,
 )
 @router.post(  # type: ignore[misc]
-    "/search", summary="搜索所有订阅", response_model=Response[None]
+    "/search",
+    summary="搜索所有订阅",
+    response_model=Response[SubscriptionSearchSubmissionSchema],
 )
 async def search_subscribes(
     command: SearchSubscriptionsCommand = Depends(get_search_subscriptions_command),
     current_user: ApiPrincipal = Depends(get_current_active_user_async),
 ) -> Any:
     """搜索当前用户可管理的全部订阅。"""
-    await command.execute(
+    submission = await command.execute(
         SubscribeSearchActor(
             username=current_user.name,
             is_superuser=current_user.is_superuser,
         )
     )
-    return Response(success=True)
+    if submission is None:
+        return Response(success=False, message="没有需要搜索的订阅")
+    return Response(
+        success=True,
+        message=_search_submission_message(submission),
+        data=_search_submission_schema(submission),
+    )
 
 
 @router.get(  # type: ignore[misc]
     "/search/{subscribe_id}",
     summary="搜索订阅（兼容入口）",
-    response_model=Response[None],
+    response_model=Response[SubscriptionSearchSubmissionSchema],
     include_in_schema=False,
     deprecated=True,
 )
 @router.post(  # type: ignore[misc]
     "/search/{subscribe_id}",
     summary="搜索订阅",
-    response_model=Response[None],
+    response_model=Response[SubscriptionSearchSubmissionSchema],
 )
 async def search_subscribe(
     subscribe_id: int,
@@ -136,13 +182,17 @@ async def search_subscribe(
     current_user: ApiPrincipal = Depends(get_current_active_user_async),
 ) -> Any:
     """根据订阅编号搜索一个订阅。"""
-    found = await command.execute(
+    submission = await command.execute(
         SubscribeSearchActor(
             username=current_user.name,
             is_superuser=current_user.is_superuser,
         ),
         subscribe_id=subscribe_id,
     )
-    if not found:
+    if submission is None:
         return Response(success=False, message="订阅不存在")
-    return Response(success=True)
+    return Response(
+        success=True,
+        message=_search_submission_message(submission),
+        data=_search_submission_schema(submission),
+    )

app/application/subscription/contract.py

@@ -596,8 +596,12 @@ class SubscriptionStagingPort(Protocol):
         """异步按媒体身份读取删除候选快照。"""
         ...
 
-    async def list_search_ids(self, username: str, state: str) -> builtins.list[int]:
-        """异步读取用户可搜索订阅主键。"""
+    async def list_search_ids(
+        self,
+        username: Optional[str],
+        state: str,
+    ) -> builtins.list[int]:
+        """异步读取用户或管理员全局范围内可搜索的订阅主键。"""
         ...
 
     async def stage_delete(self, subscribe_id: int) -> None:

app/application/subscription/execution.py

@@ -125,11 +125,13 @@ def handle_subscription_search_deferred(
     deferred: SubscriptionSearchDeferred,
     record: Callable[..., None],
 ) -> None:
-    """把站点预算冲突重新入队，并记录为可恢复而非失败的任务结果。"""
+    """把站点暂时不可用的任务重新入队，而不是记录为搜索失败。"""
     requeued = queue.defer_task(
         task_id=task_id,
         lease_token=lease_token,
         available_at=deferred.retry_at,
+        phase="waiting_site_budget",
+        message="站点暂时忙，系统会自动继续搜索",
     )
     if requeued:
         record("requeued", "site_budget_deferred")
@@ -187,6 +189,7 @@ class SearchEnqueueResult:
     batch: SearchBatchSnapshot
     created_count: int
     coalesced_count: int
+    active_batch_ids: tuple[str, ...]
 
 
 class SubscriptionSearchRepository(Protocol):
@@ -245,8 +248,10 @@ class SubscriptionSearchRepository(Protocol):
         task_id: str,
         lease_token: str,
         available_at: str,
+        phase: str = "waiting_site_budget",
+        message: Optional[str] = None,
     ) -> bool:
-        """把临时站点预算冲突任务退回队列，并设置下一次领取时间。"""
+        """把临时不可执行任务退回队列，并保留用户可理解的等待原因。"""
         ...
 
     def is_cancel_requested(self, task_id: str) -> bool:

app/application/subscription/observability.py

@@ -16,30 +16,73 @@ from app.application.subscription.sitebudget import SubscriptionSiteBudgetMetric
 
 SearchTaskOutcome = Literal["completed", "skipped", "failed", "cancelled", "requeued"]
 
+_SEARCH_SOURCE_NAMES = {
+    "manual": "手动订阅搜索",
+    "targeted": "指定订阅搜索",
+    "new": "新订阅自动搜索",
+    "fallback": "订阅定时检查",
+    "resume": "等待中的订阅搜索",
+    "inline": "订阅搜索",
+}
+
+_SEARCH_STATE_NAMES = {
+    "completed": "已完成",
+    "failed": "部分订阅没有完成",
+    "cancelled": "已停止",
+    "skipped": "部分订阅这次未搜索",
+    "queued": "部分订阅稍后继续",
+    "running": "仍在处理中",
+    "cancelling": "正在停止",
+    "stopped": "部分订阅稍后继续",
+}
+
+_MATCH_STATE_NAMES = {
+    "completed": "检查完成",
+    "failed": "部分订阅没有完成",
+    "cancelled": "已停止",
+    "skipped": "部分订阅这次未检查",
+    "stopped": "已停止，部分订阅这次未检查",
+}
+
+
+def _search_source_name(source: str) -> str:
+    """返回搜索来源对应的可读名称。"""
+    return _SEARCH_SOURCE_NAMES.get(source, "订阅搜索")
+
+
+def _search_state_name(state: str) -> str:
+    """返回搜索结束状态对应的可读说明。"""
+    return _SEARCH_STATE_NAMES.get(state, "已结束")
+
+
+def _match_state_name(state: str) -> str:
+    """返回订阅资源检查状态对应的可读说明。"""
+    return _MATCH_STATE_NAMES.get(state, "已结束")
+
 
 def batch_progress_text(batch: Optional[SearchBatchSnapshot]) -> str:
     """把批次聚合终态转为兼容进度文案。"""
     if batch is None:
-        return "订阅搜索任务已提交"
+        return "搜索已安排"
     if batch.state == "failed":
-        return "订阅搜索完成，部分任务失败"
+        return "搜索结束，部分订阅没有完成"
     if batch.state == "cancelled":
-        return "订阅搜索已取消"
+        return "搜索已停止"
     if batch.state == "skipped":
-        return "订阅搜索完成，部分任务本轮已跳过"
+        return "搜索结束，部分订阅这次未搜索"
     if batch.state in {"queued", "running", "cancelling"}:
-        return "订阅搜索任务已排队"
+        return "搜索已安排，系统正在依次处理"
     if batch.skipped_count:
-        return "订阅搜索完成，部分任务本轮已跳过"
-    return "订阅搜索完成"
+        return "搜索结束，部分订阅这次未搜索"
+    return "搜索完成"
 
 
 def inline_search_result(total: int, finished: int) -> tuple[str, dict[str, int]]:
     """返回兼容搜索的真实终态文案与计数。"""
     text = (
-        "订阅搜索完成"
+        "搜索完成"
         if finished == total
-        else "订阅搜索结束，部分订阅本轮未执行或未完成"
+        else "搜索结束，部分订阅这次没有完成"
     )
     return text, {"total": total, "finished": finished}
 
@@ -77,14 +120,14 @@ def finish_returned_search_task(
                 task_id=task_id,
                 lease_token=lease_token,
                 state="completed",
-                error="执行截止时间晚于下载提交边界，已按实际结果完成",
+                error="下载已经提交，虽然搜索用时较长，结果仍然有效",
             )
             return subscription_id, "completed", "ttl_timeout"
         queue.finish_task(
             task_id=task_id,
             lease_token=lease_token,
             state="failed",
-            error="订阅执行已超过协作截止时间",
+            error="这次搜索用时过长，已停止，可稍后重试",
         )
         return None, "failed", "ttl_timeout"
     if system_stopped:
@@ -93,7 +136,7 @@ def finish_returned_search_task(
                 task_id=task_id,
                 lease_token=lease_token,
                 state="completed",
-                error="停机请求晚于下载提交边界，已按实际结果完成",
+                error="下载已经提交，系统停止后仍保留这次结果",
             )
             return subscription_id, "completed", "system_stop"
         queue.release_task(task_id=task_id, lease_token=lease_token)
@@ -104,7 +147,7 @@ def finish_returned_search_task(
                 task_id=task_id,
                 lease_token=lease_token,
                 state="completed",
-                error="取消请求晚于下载提交边界，已按实际结果完成",
+                error="下载已经提交，无法撤回，已保留这次结果",
             )
             return subscription_id, "completed", "cancelled"
         queue.release_task(task_id=task_id, lease_token=lease_token, cancelled=True)
@@ -173,26 +216,31 @@ class MatchExecutionSummary:
         return "completed"
 
     def start_log(self) -> str:
-        """构造 Match 轮次开始日志。"""
+        """构造不暴露内部标记的资源检查开始日志。"""
         return (
-            "订阅治理轮次开始: operation=match "
-            f"run_id={self.run_id} subscriptions={self.total} "
-            f"sites={self.site_count} candidates={self.candidate_count}"
+            f"开始检查订阅资源，共 {self.total} 个订阅、{self.candidate_count} 个资源，"
+            f"来自 {self.site_count} 个站点。"
         )
 
     def finish_log(self) -> str:
-        """构造 Match 轮次结束日志，completed 表示任务处理完成而非订阅成功。"""
-        elapsed_ms = (time.perf_counter() - self.started_at) * 1000
-        return (
-            "订阅治理轮次结束: operation=match "
-            f"run_id={self.run_id} state={self.state} subscriptions={self.total} "
-            f"processed={self.finished} task_completed={self.completed} "
-            f"task_skipped={self.skipped} task_failed={self.failed} "
-            f"admission_conflicts={self.admission_conflicts} cancelled={self.cancelled} "
-            f"ttl_timeouts={self.ttl_timeouts} release_failures={self.release_failures} "
-            f"sites={self.site_count} candidates={self.candidate_count} "
-            f"duration_ms={elapsed_ms:.1f}"
+        """构造可直接阅读的资源检查结束日志。"""
+        elapsed_seconds = time.perf_counter() - self.started_at
+        message = (
+            f"订阅资源检查结束：{_match_state_name(self.state)}。"
+            f"本次检查 {self.finished}/{self.total} 个订阅，完成 {self.completed} 个，"
+            f"这次未检查 {self.skipped} 个，失败 {self.failed} 个；"
+            f"共检查 {self.candidate_count} 个资源，来自 {self.site_count} 个站点，"
+            f"用时 {elapsed_seconds:.1f} 秒。"
         )
+        if self.admission_conflicts:
+            message += f"其中 {self.admission_conflicts} 个订阅正在处理中，本次没有重复检查。"
+        if self.cancelled:
+            message += f"另有 {self.cancelled} 个订阅已停止。"
+        if self.ttl_timeouts:
+            message += f"另有 {self.ttl_timeouts} 个订阅检查时间过长，已停止。"
+        if self.release_failures:
+            message += f"另有 {self.release_failures} 个订阅的搜索状态没有正常恢复，系统稍后会继续检查。"
+        return message
 
     def as_data(self, current: Optional[int] = None) -> dict[str, int]:
         """构造供进度消费者读取的实际计数。"""
@@ -256,15 +304,14 @@ class SearchExecutionSummary:
             self.ttl_timeouts += 1
 
     def start_log(self) -> str:
-        """构造 Search 轮次开始日志。"""
-        return (
-            "订阅治理轮次开始: operation=search "
-            f"run_id={self.run_id} batch_id={self.batch_id or '-'} source={self.source} "
-            f"subscriptions={self.requested} coalesced={self.coalesced}"
-        )
+        """构造不暴露内部术语的 Search 开始日志。"""
+        message = f"开始{_search_source_name(self.source)}，共 {self.requested} 个订阅"
+        if self.coalesced:
+            message += f"，其中 {self.coalesced} 个已经在处理中"
+        return f"{message}。"
 
     def finish_log(self, batch: Optional[SearchBatchSnapshot] = None) -> str:
-        """构造 Search 轮次结束日志，任务完成与订阅完成保持分离。"""
+        """构造可直接阅读的 Search 结束日志。"""
         sites = self.site_metrics.snapshot()
         if self.round_failed:
             state = "failed"
@@ -282,19 +329,16 @@ class SearchExecutionSummary:
             state = "skipped"
         else:
             state = "completed"
-        elapsed_ms = (time.perf_counter() - self.started_at) * 1000
-        return (
-            "订阅治理轮次结束: operation=search "
-            f"run_id={self.run_id} batch_id={self.batch_id or '-'} source={self.source} "
-            f"state={state} subscriptions={self.requested} processed={self.processed} "
-            f"task_completed={self.completed} task_skipped={self.skipped} "
-            f"task_failed={self.failed} task_cancelled={self.cancelled} task_requeued={self.requeued} "
-            f"admission_conflicts={self.admission_conflicts} cancelled={self.cancelled} "
-            f"ttl_timeouts={self.ttl_timeouts} consumer_conflicts={self.consumer_conflicts} "
-            f"round_failed={int(self.round_failed)} "
-            f"sites={sites.site_count} site_requests={sites.request_count} "
-            f"site_failures={sites.failure_count} site_cooldown_skips={sites.cooldown_skip_count} "
-            f"candidates={sites.candidate_count} cooldown_seconds={sites.cooldown_seconds:.1f} "
-            f"release_failures={self.release_failures + sites.release_failure_count} "
-            f"duration_ms={elapsed_ms:.1f}"
+        elapsed_seconds = time.perf_counter() - self.started_at
+        message = (
+            f"{_search_source_name(self.source)}结束：{_search_state_name(state)}。"
+            f"本次处理 {self.processed}/{self.requested} 个，完成 {self.completed} 个，"
+            f"这次未搜索 {self.skipped} 个，失败 {self.failed} 个，"
+            f"稍后继续 {self.requeued} 个，已停止 {self.cancelled} 个；"
+            f"访问 {sites.site_count} 个站点，发出 {sites.request_count} 次请求，"
+            f"找到 {sites.candidate_count} 个资源，用时 {elapsed_seconds:.1f} 秒。"
         )
+        unfinished_cleanup = self.release_failures + sites.release_failure_count
+        if unfinished_cleanup:
+            message += f"另有 {unfinished_cleanup} 个搜索状态没有正常恢复，系统稍后会继续处理。"
+        return message

app/application/subscription/search.py

@@ -1,7 +1,8 @@
 """手工订阅搜索应用用例。"""
 
+from collections.abc import Awaitable, Callable
 from dataclasses import dataclass
-from typing import Callable
+from typing import Optional
 
 from app.application.subscription.contract import (
     SubscribeDeletionCandidate,
@@ -17,45 +18,65 @@ class SubscribeSearchActor:
     is_superuser: bool
 
 
-SubscribeSearchScheduler = Callable[[tuple[int, ...] | None, str | None], None]
+@dataclass(frozen=True, slots=True)
+class SubscriptionSearchSubmission:
+    """一次手工搜索请求的入队结果。"""
+
+    batch_ids: tuple[str, ...]
+    target_count: int
+    queued_count: int
+    ongoing_count: int
+    single: bool
+
+    @property
+    def batch_id(self) -> Optional[str]:
+        """返回最适合前端立即跟踪的批次编号。"""
+        return self.batch_ids[0] if self.batch_ids else None
+
+
+SubscribeSearchSubmitter = Callable[
+    [tuple[int, ...], bool],
+    Awaitable[SubscriptionSearchSubmission],
+]
 
 
 class SearchSubscriptionsCommand:
-    """按用户权限生成并提交手工订阅搜索任务。"""
+    """按用户权限读取目标，并把手工搜索轻量提交到持久队列。"""
 
     def __init__(
         self,
         repository: SubscriptionStagingPort,
-        schedule_search: SubscribeSearchScheduler,
+        submit_search: SubscribeSearchSubmitter,
     ) -> None:
-        """注入订阅读取端口和后台任务提交端口。"""
+        """注入订阅读取端口和持久队列提交端口。"""
         self._repository = repository
-        self._schedule_search = schedule_search
+        self._submit_search = submit_search
 
     async def execute(
         self,
         actor: SubscribeSearchActor,
         subscribe_id: int | None = None,
-    ) -> bool:
-        """提交单条或当前用户全部可搜索订阅，返回目标是否存在。"""
+    ) -> Optional[SubscriptionSearchSubmission]:
+        """提交单条或当前用户全部可搜索订阅，目标不可访问时返回空。"""
         if subscribe_id is not None:
             candidate = await self._repository.get_candidate(subscribe_id)
             if not self._can_access(candidate, actor):
-                return False
-            self._schedule_search((subscribe_id,), None)
-            return True
-
-        if actor.is_superuser:
-            self._schedule_search(None, "R")
-            return True
+                return None
+            return await self._submit_search((subscribe_id,), True)
 
         subscribe_ids = await self._repository.list_search_ids(
-            actor.username,
+            None if actor.is_superuser else actor.username,
             "R",
         )
-        if subscribe_ids:
-            self._schedule_search(tuple(subscribe_ids), None)
-        return True
+        if not subscribe_ids:
+            return SubscriptionSearchSubmission(
+                batch_ids=(),
+                target_count=0,
+                queued_count=0,
+                ongoing_count=0,
+                single=False,
+            )
+        return await self._submit_search(tuple(subscribe_ids), False)
 
     @staticmethod
     def _can_access(

app/application/subscription/sitebudget.py

@@ -27,20 +27,28 @@ class SubscriptionSearchDeferred(RuntimeError):
     """表示订阅搜索未失败，而是应在站点预算可用后重新入队。"""
 
     def __init__(self, *, retry_at: str, site_ids: tuple[int, ...]) -> None:
-        """保存队列恢复所需的时间和冲突站点，避免把临时冲突写成错误。"""
-        super().__init__(f"订阅搜索已延后，站点预算最早可重试：{retry_at}")
+        """保存队列恢复所需的时间和站点，避免把临时等待写成错误。"""
+        super().__init__("站点暂时忙，系统会自动继续搜索")
         self.retry_at = retry_at
         self.site_ids = site_ids
 
 
 class SubscriptionSiteBudgetUnavailable(RuntimeError):
-    """表示站点预算暂时不可用，调用方应记录为延后而非失败。"""
+    """表示站点暂时不可用，调用方应记录为等待而非失败。"""
 
-    def __init__(self, *, site_id: int, retry_at: str) -> None:
+    def __init__(
+        self,
+        *,
+        site_id: int,
+        retry_at: str,
+        wait_reason: Optional[str] = None,
+    ) -> None:
         """保存站点和下一次可尝试时间，供订阅队列恢复。"""
-        super().__init__(f"站点 {site_id} 冷却或已有在途搜索，最早可重试：{retry_at}")
+        message = "站点正在处理其他搜索，稍后会自动继续" if wait_reason == "busy" else "站点暂时不可用，稍后会自动重试"
+        super().__init__(message)
         self.site_id = site_id
         self.retry_at = retry_at
+        self.wait_reason = wait_reason
 
 
 @dataclass(frozen=True, slots=True)
@@ -52,6 +60,7 @@ class SiteBudgetClaim:
     retry_at: str
     consecutive_failures: int
     lease_token: Optional[str] = None
+    wait_reason: Optional[str] = None
 
 
 @dataclass(frozen=True, slots=True)
@@ -195,6 +204,7 @@ class SubscriptionSiteBudget:
         raise SubscriptionSiteBudgetUnavailable(
             site_id=site_id,
             retry_at=claim.retry_at,
+            wait_reason=claim.wait_reason,
         )
 
     def _report_phase(self, phase: str, site_id: Optional[int]) -> None:
@@ -251,4 +261,4 @@ class SubscriptionSiteBudget:
     def _raise_if_cancelled(self) -> None:
         """在创建站点租约前传播取消或停机。"""
         if self._stop_state.is_system_stopped or self._cancelled():
-            raise SubscriptionSearchCancelled("订阅搜索已取消")
+            raise SubscriptionSearchCancelled("搜索已停止")

app/application/subscription/status.py

@@ -18,6 +18,7 @@ class SubscriptionExecutionStatus:
     batch_id: Optional[str] = None
     task_id: Optional[str] = None
     current_site_id: Optional[int] = None
+    next_run_at: Optional[str] = None
     error: Optional[str] = None
     can_cancel: bool = False
 
@@ -79,6 +80,8 @@ class SubscriptionExecutionStatusService:
         "running",
         "matching",
         "searching",
+        "scheduled",
+        "waiting_subscription",
         "waiting_site_budget",
         "preparing",
         "submitting",
@@ -153,8 +156,8 @@ class SubscriptionExecutionStatusService:
             state = phase = "cancelling"
         elif task.state == "running":
             state = phase = task.phase or "running"
-        elif task.state == "queued" and task.phase == "waiting_site_budget":
-            state = phase = "waiting_site_budget"
+        elif task.state == "queued" and task.phase in cls._ACTIVE_STATES:
+            state = phase = task.phase
         else:
             state = phase = task.state
         return SubscriptionExecutionStatus(
@@ -164,6 +167,7 @@ class SubscriptionExecutionStatusService:
             batch_id=task.batch_id,
             task_id=task.task_id,
             current_site_id=task.current_site_id,
+            next_run_at=task.available_at if task.state == "queued" else None,
             updated_at=task.updated_at,
             error=cls._safe_error(task.last_error),
             can_cancel=state in cls._ACTIVE_STATES,

app/chain/subscribe/contract.py

@@ -50,6 +50,7 @@ if TYPE_CHECKING:
         _SubscribeChain__is_full_season_best_version_resource: Callable[..., Any]
         _SubscribeChain__notify_subscribe_create_failure: Callable[..., Any]
         _SubscribeChain__post_subscribe_added: Callable[..., Any]
+        _SubscribeChain__queue_new_subscription_search: Callable[..., Any]
         _SubscribeChain__prepare_best_version_tv_candidate: Callable[..., Any]
         _SubscribeChain__prepare_subscribe_progress_fields: Callable[..., Any]
         _SubscribeChain__prepare_total_episode_change_fields: Callable[..., Any]
@@ -71,7 +72,7 @@ if TYPE_CHECKING:
         _report_search_progress: Callable[..., Any]
         _search_music_subscribe: Callable[..., Any]
         _subscription_query: Callable[..., Any]
-        _defer_recent_subscription: Callable[..., Any]
+        _recent_subscription_retry_at: Callable[..., Any]
         _validate_music_subscribe_target: Callable[..., Any]
         _wait_before_scheduled_search: Callable[..., Any]
         add: Callable[..., Any]

app/chain/subscribe/match.py

@@ -85,14 +85,11 @@ def _release_match_admission(
     subscription_id: int,
     summary: MatchExecutionSummary,
 ) -> None:
-    """释放 Match 订阅准入，并让 token 冲突即时可见。"""
+    """结束当前订阅资源检查，并记录未能恢复的状态。"""
     if admission.release(lease):
         return
     summary.release_failures += 1
-    logger.error(
-        "订阅准入释放失败: operation=match "
-        f"subscription_id={subscription_id} run_id={summary.run_id}"
-    )
+    logger.error(f"订阅 {subscription_id} 的搜索状态没有正常恢复，系统稍后会继续检查")
 
 
 def _report_match_finished(
@@ -102,13 +99,13 @@ def _report_match_finished(
     """按实际计数发布 Match 最终进度。"""
     if not progress_callback:
         return
-    final_text = "订阅资源匹配完成"
+    final_text = "订阅资源检查完成"
     if summary.finished < summary.total:
-        final_text = "订阅资源匹配已停止，部分订阅未执行"
+        final_text = "订阅资源检查已停止，部分订阅这次未检查"
     elif summary.failed:
-        final_text = "订阅资源匹配完成，部分订阅失败"
+        final_text = "订阅资源检查结束，部分订阅没有完成"
     elif summary.skipped:
-        final_text = "订阅资源匹配完成，部分订阅跳过"
+        final_text = "订阅资源检查完成，部分订阅这次未检查"
     _report_match_progress(progress_callback, summary, value=100, text=final_text)
 
 
@@ -234,7 +231,7 @@ def _prepare_subscription_match(
     try:
         meta = build_subscribe_meta(subscribe)
     except ValueError:
-        logger.error(f"订阅 {subscribe.name} 类型错误：{subscribe.type}")
+        logger.error(f"订阅《{subscribe.name}》的媒体类型不受支持，暂时无法检查资源")
         return None
     domains = owner.site_repository.get_domains_by_ids(subscribe.sites) if subscribe.sites else []
     sub_sites = owner.get_sub_sites(subscribe)
@@ -337,12 +334,12 @@ class SubscribeMatchOwner(_SubscribeOwnerBase):
         该入口保持订阅刷新、定时任务和插件调用的稳定签名，具体匹配流程由内部阶段执行。
         """
         if not torrents:
-            logger.warn("没有缓存资源，无法匹配订阅")
+            logger.warn("当前没有可检查的订阅资源")
             if progress_callback:
-                progress_callback(value=100, text="没有缓存资源，跳过订阅匹配")
+                progress_callback(value=100, text="当前没有可检查的订阅资源")
             return
         if progress_callback:
-            progress_callback(value=0, text="正在预处理订阅资源 ...")
+            progress_callback(value=0, text="正在整理订阅资源 ...")
         return self._run_match(
             torrents=torrents,
             progress_callback=progress_callback,
@@ -360,13 +357,13 @@ class SubscribeMatchOwner(_SubscribeOwnerBase):
         :param progress_callback: 订阅匹配进度更新回调
         """
         if not torrents:
-            logger.warn("没有缓存资源，无法匹配订阅")
+            logger.warn("当前没有可检查的订阅资源")
             if progress_callback:
-                progress_callback(value=100, text="没有缓存资源，跳过订阅匹配")
+                progress_callback(value=100, text="当前没有可检查的订阅资源")
             return
 
         if progress_callback:
-            progress_callback(value=0, text="正在预处理订阅资源 ...")
+            progress_callback(value=0, text="正在整理订阅资源 ...")
 
         lock_acquired = False
         summary = MatchExecutionSummary.from_candidates(torrents)
@@ -390,7 +387,7 @@ class SubscribeMatchOwner(_SubscribeOwnerBase):
                     progress_callback,
                     summary,
                     value=20,
-                    text=f"资源预处理完成，开始匹配 {total_num} 个订阅 ...",
+                    text=f"资源整理完成，开始检查 {total_num} 个订阅 ...",
                 )
             try:
                 for index, listed_subscribe in enumerate(subscribes, start=1):
@@ -401,7 +398,7 @@ class SubscribeMatchOwner(_SubscribeOwnerBase):
                             progress_callback,
                             summary,
                             value=20 + ((index - 1) / total_num * 80 if total_num else 80),
-                            text=(f"正在匹配订阅（{index}/{total_num}）{listed_subscribe.name} ..."),
+                            text=(f"正在检查订阅（{index}/{total_num}）{listed_subscribe.name} ..."),
                             current=listed_subscribe.id,
                         )
                     outcome: Optional[str] = "skipped"
@@ -413,7 +410,7 @@ class SubscribeMatchOwner(_SubscribeOwnerBase):
                     )
                     if lease is None:
                         reason = "admission_conflict"
-                        logger.debug(f"订阅 {listed_subscribe.name} 正在由其他通道处理，本轮匹配已跳过")
+                        logger.debug(f"订阅 {listed_subscribe.name} 正在处理中，本次不再重复检查资源")
                     else:
                         current_subscribe = None
                         execution_context = SubscriptionExecutionContext(
@@ -424,14 +421,12 @@ class SubscribeMatchOwner(_SubscribeOwnerBase):
                         try:
                             current_subscribe = self.subscription_repository.get(listed_subscribe.id)
                             if current_subscribe is None:
-                                logger.debug(f"订阅 {listed_subscribe.id} 已不存在，本轮匹配跳过")
+                                logger.debug(f"订阅 {listed_subscribe.id} 已删除，本次不再检查资源")
                             elif current_subscribe.state not in {"R", "P"}:
-                                logger.debug(
-                                    f"订阅 {current_subscribe.name} 当前状态为 {current_subscribe.state}，本轮匹配跳过"
-                                )
+                                logger.debug(f"订阅 {current_subscribe.name} 当前不需要检查资源，本次跳过")
                             elif execution_context.should_stop():
                                 reason = _match_stop_reason(execution_context)
-                                logger.debug(f"订阅 {current_subscribe.name} 已取消或超时，本轮匹配跳过")
+                                logger.debug(f"订阅 {current_subscribe.name} 的资源检查已停止")
                             else:
                                 outcome = self._match_subscription(
                                     subscribe=current_subscribe,
@@ -444,14 +439,14 @@ class SubscribeMatchOwner(_SubscribeOwnerBase):
                         except SubscriptionSearchCancelled as err:
                             outcome = "skipped"
                             reason = _match_stop_reason(execution_context) or "cancelled"
-                            logger.debug(f"订阅 {listed_subscribe.name} 匹配已取消：{str(err)}")
+                            logger.debug(f"订阅 {listed_subscribe.name} 的资源检查已停止：{str(err)}")
                         except Exception as err:
                             outcome = "failed"
                             reason = "error"
                             subscribe_name = (
                                 current_subscribe.name if current_subscribe is not None else listed_subscribe.name
                             )
-                            logger.error(f"订阅 {subscribe_name} 匹配失败：{str(err)}", exc_info=True)
+                            logger.error(f"订阅 {subscribe_name} 检查资源时出错：{str(err)}", exc_info=True)
                         finally:
                             _release_match_admission(
                                 self._subscription_execution_admission,
@@ -464,7 +459,7 @@ class SubscribeMatchOwner(_SubscribeOwnerBase):
                         progress_callback,
                         summary,
                         value=20 + (summary.finished / total_num * 80 if total_num else 80),
-                        text=(f"订阅匹配（{index}/{total_num}）处理完成"),
+                        text=(f"已检查订阅（{index}/{total_num}）"),
                     )
             finally:
                 processed_torrents.clear()
@@ -477,7 +472,7 @@ class SubscribeMatchOwner(_SubscribeOwnerBase):
                 logger.info(summary.finish_log())
             if lock_acquired:
                 self._match_lock.release()
-                logger.debug(f"match Lock released at {datetime.now()}")
+                logger.debug(f"订阅资源检查已结束：{datetime.now()}")
 
     def _match_subscription(
         self,

app/chain/subscribe/notify.py

@@ -18,6 +18,7 @@ from app.domain.context import (
     MediaInfo,
 )
 from app.domain.meta.metabase import MetaBase
+from app.runtime.execution import run_in_threadpool
 from app.runtime.log import logger
 from app.schemas.common import JsonData
 from app.schemas.message import Message as _SchemaMessage
@@ -142,6 +143,13 @@ class SubscribeNotificationOwner(_SubscribeOwnerBase):
                 "mediainfo": context.mediainfo.to_dict(),
             },
         )
+        try:
+            self._SubscribeChain__queue_new_subscription_search(subscribe_id)
+        except Exception as error:
+            logger.warning(
+                "订阅已保存，但自动搜索暂时没有安排成功，"
+                f"系统会在下一次检查时重试：{error}"
+            )
         try:
             report_delivered = _subscription_share_snapshot().report_added(
                 self._SubscribeChain__subscribe_report_payload(context)
@@ -177,6 +185,16 @@ class SubscribeNotificationOwner(_SubscribeOwnerBase):
                 "mediainfo": context.mediainfo.to_dict(),
             },
         )
+        try:
+            await run_in_threadpool(
+                self._SubscribeChain__queue_new_subscription_search,
+                subscribe_id,
+            )
+        except Exception as error:
+            logger.warning(
+                "订阅已保存，但自动搜索暂时没有安排成功，"
+                f"系统会在下一次检查时重试：{error}"
+            )
         try:
             report_delivered = await _subscription_share_snapshot().async_report_added(
                 self._SubscribeChain__subscribe_report_payload(context)

app/chain/subscribe/search.py

@@ -53,6 +53,10 @@ from app.schemas.types import (
     SystemConfigKey,
 )
 
+_NEW_SUBSCRIPTION_EDIT_SECONDS = 60
+_FOREGROUND_RETRY_SECONDS = 5
+_BACKGROUND_RETRY_SECONDS = 10
+
 
 def _ensure_execution_active(
     execution_context: Optional[SubscriptionExecutionContext],
@@ -61,9 +65,9 @@ def _ensure_execution_active(
     if execution_context is None:
         return
     if execution_context.is_cancel_requested():
-        raise SubscriptionSearchCancelled("订阅搜索已取消")
+        raise SubscriptionSearchCancelled("搜索已停止")
     if execution_context.is_expired():
-        raise TimeoutError("订阅执行已超过协作截止时间")
+        raise TimeoutError("这次搜索用时过长，已停止")
 
 
 def _update_search_task_phase(
@@ -91,7 +95,8 @@ def _search_source_and_priority(
 ) -> tuple[str, int]:
     """把兼容入口归一为持久来源和公平队列优先级。"""
     if manual:
-        return "manual", 100
+        target_count = 1 if sid else len(sids or ())
+        return "manual", 120 if target_count == 1 else 100
     if sid or sids is not None:
         return "targeted", 80
     if state in {"R", "P"}:
@@ -99,6 +104,13 @@ def _search_source_and_priority(
     return "new", 50
 
 
+def _retry_at_after(seconds: int) -> str:
+    """返回指定秒数后的 UTC 时间，供短暂等待任务重新入队。"""
+    return (datetime.now(timezone.utc) + timedelta(seconds=max(1, seconds))).isoformat(
+        timespec="seconds"
+    )
+
+
 def _search_task_available_at(
     source: str,
     subscription_ids: tuple[int, ...],
@@ -179,16 +191,16 @@ class _SubscribeSearchQueueCoordinator(_SubscribeOwnerBase):
         progress_callback: Optional[Callable[..., None]],
     ) -> bool:
         """获取 Search 或 Match 通道锁，保持各通道内部串行。"""
+        operation_label = {"search": "搜索", "match": "资源检查"}[operation]
         lock = self._match_lock if operation == "match" else self._search_queue_lock
         if lock.acquire(blocking=True, timeout=self._SUBSCRIPTION_EXECUTION_TTL):
-            logger.debug(f"{operation} lock acquired at {datetime.now()}")
+            logger.debug(f"订阅{operation_label}已开始：{datetime.now()}")
             return True
-        operation_label = {"search": "搜索", "match": "匹配"}[operation]
         progress_text = {
-            "search": "订阅搜索锁等待超时，已跳过本轮",
-            "match": "订阅匹配锁等待超时，已跳过本轮",
+            "search": "订阅搜索正在处理中，本次不再重复开始",
+            "match": "订阅资源检查正在进行，本次不再重复开始",
         }[operation]
-        logger.error(f"订阅{operation_label}锁等待超时，已中止本轮执行")
+        logger.error(f"订阅{operation_label}等待时间过长，本次不再重复开始")
         if progress_callback:
             progress_callback(
                 value=100,
@@ -273,6 +285,12 @@ class _SubscribeSearchQueueCoordinator(_SubscribeOwnerBase):
         try:
             subscribes = self._load_search_subscriptions(sid=sid, sids=sids, state=state)
             total = len(subscribes)
+            source, _priority = _search_source_and_priority(
+                sid=sid,
+                sids=sids,
+                state=state,
+                manual=manual,
+            )
             summary = SearchExecutionSummary(source="inline", requested=total)
             logger.info(summary.start_log())
             if progress_callback:
@@ -287,7 +305,7 @@ class _SubscribeSearchQueueCoordinator(_SubscribeOwnerBase):
                     summary.stopped = True
                     break
                 self._report_search_progress(progress_callback, subscribe, index, total)
-                if self._defer_recent_subscription(subscribe):
+                if self._recent_subscription_retry_at(subscribe, source):
                     summary.record("skipped", "recent_subscription")
                     continue
                 self._wait_before_scheduled_search(sid, sids, state, progress_callback)
@@ -297,7 +315,7 @@ class _SubscribeSearchQueueCoordinator(_SubscribeOwnerBase):
                     ttl_seconds=self._SUBSCRIPTION_EXECUTION_TTL,
                 )
                 if lease is None:
-                    logger.debug(f"订阅 {subscribe.name} 正在由其他通道处理，本轮搜索已跳过")
+                    logger.debug(f"订阅《{subscribe.name}》正在处理，本次搜索先不重复执行")
                     summary.record("skipped", "admission_conflict")
                     continue
                 execution_context = SubscriptionExecutionContext(
@@ -312,7 +330,7 @@ class _SubscribeSearchQueueCoordinator(_SubscribeOwnerBase):
                     current = self.subscription_repository.get(subscribe.id)
                     if current is None or current.state == "S":
                         if current and current.state == "S":
-                            logger.debug(f"订阅 {current.name} 已暂停，本轮搜索已跳过")
+                            logger.debug(f"订阅《{current.name}》已暂停，本次没有搜索")
                         continue
                     processed_result = self._process_search_subscription(
                         current,
@@ -326,12 +344,12 @@ class _SubscribeSearchQueueCoordinator(_SubscribeOwnerBase):
                 except SubscriptionSearchCancelled:
                     reason = "ttl_timeout" if execution_context.is_expired() else "cancelled"
                     outcome = "failed" if reason == "ttl_timeout" else "cancelled"
-                    logger.debug(f"订阅 {subscribe.name} 搜索已在安全边界取消")
+                    logger.debug(f"订阅《{subscribe.name}》搜索已停止")
                 except SubscriptionSearchDeferred as deferred:
                     outcome = "skipped"
                     reason = "site_budget_deferred"
                     logger.debug(
-                        f"订阅 {subscribe.name} 站点预算冲突，兼容搜索将在 {deferred.retry_at} 后重试"
+                        f"订阅《{subscribe.name}》遇到站点繁忙，将在稍后自动重试：{deferred.retry_at}"
                     )
                 except Exception as err:
                     outcome = "failed"
@@ -347,7 +365,8 @@ class _SubscribeSearchQueueCoordinator(_SubscribeOwnerBase):
                             )
                     except Exception as err:
                         logger.error(
-                            f"订阅 {subscribe.name} 搜索后状态重置失败：{str(err)}",
+                            f"订阅《{subscribe.name}》搜索结束后没有恢复到正常状态，"
+                            f"系统稍后会继续处理：{str(err)}",
                             exc_info=True,
                         )
                     finally:
@@ -355,8 +374,7 @@ class _SubscribeSearchQueueCoordinator(_SubscribeOwnerBase):
                         if not released:
                             summary.release_failures += 1
                             logger.error(
-                                "订阅准入释放失败: operation=search "
-                                f"subscription_id={subscribe.id} run_id={summary.run_id}"
+                                f"订阅《{subscribe.name}》的搜索状态没有正常恢复，系统稍后会继续处理"
                             )
                         summary.record(outcome, reason)
                         self._report_search_progress(
@@ -375,7 +393,7 @@ class _SubscribeSearchQueueCoordinator(_SubscribeOwnerBase):
             self._search_queue_lock.release()
             if summary is not None:
                 logger.info(summary.finish_log())
-            logger.debug(f"search Lock released at {datetime.now()}")
+            logger.debug(f"订阅搜索已结束：{datetime.now()}")
 
     def _execute_queued_search(
         self,
@@ -463,7 +481,7 @@ class _SubscribeSearchQueueCoordinator(_SubscribeOwnerBase):
     ) -> set[int]:
         """有界消费可恢复任务；单任务失败不得阻止后续订阅。"""
         if not self._search_queue_lock.acquire(blocking=False):
-            logger.debug("订阅搜索队列已有消费者，本轮仅保留持久任务")
+            logger.debug("订阅搜索已经在后台进行，本次安排会接着处理")
             summary.consumer_conflicts += 1
             return set()
         owner = f"subscribe-search:{uuid4().hex}"
@@ -511,7 +529,7 @@ class _SubscribeSearchQueueOwner(_SubscribeSearchQueueCoordinator):
     ) -> Optional[int]:
         """执行一条已认领任务，返回实际进入搜索处理的订阅 ID。"""
         if task.lease_token is None:
-            logger.error(f"订阅搜索任务 {task.task_id} 缺少租约令牌，跳过执行")
+            logger.error("这次订阅搜索暂时无法开始，系统稍后会重新处理")
             summary.record("failed", "missing_lease")
             return None
         task_id = str(task.task_id)
@@ -528,14 +546,21 @@ class _SubscribeSearchQueueOwner(_SubscribeSearchQueueCoordinator):
                 task_id=task_id,
                 lease_token=lease_token,
                 state="cancelled",
-                error="订阅已不存在",
+                error="订阅已删除",
             )
             summary.record("cancelled", "missing_subscription")
             return None
         self._report_search_progress(progress_callback, subscribe, index, limit)
-        if self._defer_recent_subscription(subscribe):
-            _skip_search_task(queue, task, "订阅仍在新增保护期，本轮搜索已跳过")
-            summary.record("skipped", "recent_subscription")
+        recent_retry_at = self._recent_subscription_retry_at(subscribe, task.source)
+        if recent_retry_at:
+            queue.defer_task(
+                task_id=task_id,
+                lease_token=lease_token,
+                available_at=recent_retry_at,
+                phase="scheduled",
+                message="订阅刚刚创建，保存好设置后会自动开始搜索",
+            )
+            summary.record("requeued", "recent_subscription")
             return None
         lease = self._subscription_execution_admission.try_acquire(
             subscription_id=subscribe.id,
@@ -543,8 +568,27 @@ class _SubscribeSearchQueueOwner(_SubscribeSearchQueueCoordinator):
             ttl_seconds=self._SUBSCRIPTION_EXECUTION_TTL,
         )
         if lease is None:
-            logger.debug(f"订阅 {subscribe.name} 正在由其他通道处理，本轮搜索已跳过")
-            _skip_search_task(queue, task, "同一订阅正在由其他通道处理，本轮搜索已跳过")
+            if task.source in {"manual", "targeted", "new"}:
+                retry_seconds = (
+                    _FOREGROUND_RETRY_SECONDS
+                    if task.source in {"manual", "targeted"}
+                    else _BACKGROUND_RETRY_SECONDS
+                )
+                queue.defer_task(
+                    task_id=task_id,
+                    lease_token=lease_token,
+                    available_at=_retry_at_after(retry_seconds),
+                    phase="waiting_subscription",
+                    message="这个订阅正在处理，结束后会自动继续搜索",
+                )
+                summary.record("requeued", "admission_conflict")
+                logger.debug(f"订阅《{subscribe.name}》正在处理，搜索会在结束后自动继续")
+            else:
+                _skip_search_task(
+                    queue,
+                    task,
+                    "这个订阅正在处理，本次自动检查无需重复执行",
+                )
                 summary.record("skipped", "admission_conflict")
             return None
         phase_changed = partial(_update_search_task_phase, queue, task_id, lease_token)
@@ -563,12 +607,12 @@ class _SubscribeSearchQueueOwner(_SubscribeSearchQueueCoordinator):
                     task_id=task_id,
                     lease_token=lease_token,
                     state="cancelled",
-                    error="订阅已不存在",
+                    error="订阅已删除",
                 )
                 summary.record("cancelled", "missing_subscription")
                 return None
             if current.state == "S":
-                _skip_search_task(queue, task, "订阅已暂停，本轮搜索已跳过")
+                _skip_search_task(queue, task, "订阅已暂停，这次没有搜索")
                 summary.record("skipped", "paused")
                 return None
             searchchain.configure_subscription_site_budget(
@@ -605,7 +649,7 @@ class _SubscribeSearchQueueOwner(_SubscribeSearchQueueCoordinator):
                     task_id=task_id,
                     lease_token=lease_token,
                     state="failed",
-                    error="订阅执行已超过协作截止时间",
+                    error="这次搜索用时过长，已停止，可稍后重试",
                 )
                 summary.record("failed", "ttl_timeout")
             else:
@@ -664,7 +708,11 @@ class _SubscribeSearchQueueOwner(_SubscribeSearchQueueCoordinator):
         try:
             searchchain.configure_subscription_site_budget(None)
         except Exception as err:
-            logger.error(f"订阅 {subscribe.name} 搜索站点预算清理失败：{str(err)}", exc_info=True)
+            logger.error(
+                f"订阅《{subscribe.name}》结束站点访问时遇到问题，"
+                f"系统稍后会继续处理：{str(err)}",
+                exc_info=True,
+            )
         try:
             if current and current.state == "N":
                 self._SubscribeChain__apply_subscribe_update(
@@ -673,15 +721,12 @@ class _SubscribeSearchQueueOwner(_SubscribeSearchQueueCoordinator):
                     scene="search_reset",
                 )
         except Exception as err:
-            logger.error(f"订阅 {subscribe.name} 搜索后状态重置失败：{str(err)}", exc_info=True)
+            logger.error(f"订阅《{subscribe.name}》搜索结束后未能恢复正常状态：{str(err)}", exc_info=True)
         finally:
             released = self._subscription_execution_admission.release(lease)
             if not released:
                 summary.release_failures += 1
-                logger.error(
-                    "订阅准入释放失败: operation=search "
-                    f"subscription_id={subscribe.id} run_id={summary.run_id}"
-                )
+                logger.error(f"订阅《{subscribe.name}》的搜索状态没有正常恢复，系统稍后会继续处理")
             self._report_search_progress(
                 progress_callback,
                 subscribe,
@@ -694,26 +739,49 @@ class _SubscribeSearchQueueOwner(_SubscribeSearchQueueCoordinator):
         self,
         progress_callback: Optional[Callable[..., None]] = None,
         limit: int = 50,
+        manual_sids: Optional[tuple[int, ...]] = None,
     ) -> None:
-        """短周期恢复排队或租约过期任务，不创建新的 24 小时兜底批次。"""
+        """短周期恢复等待任务，并可优先反馈本次手工搜索结果。"""
         queue: Optional[SubscriptionSearchRepository] = getattr(
             self, "subscription_search_repository", None
         )
         if queue is None:
             return
-        summary = SearchExecutionSummary(source="resume", requested=max(1, limit))
+        manual_ids = tuple(dict.fromkeys(manual_sids or ()))
+        subscribes = (
+            self._load_search_subscriptions(sid=None, sids=manual_ids, state=None)
+            if manual_ids
+            else []
+        )
+        summary = SearchExecutionSummary(
+            source="manual" if manual_ids else "resume",
+            requested=len(manual_ids) if manual_ids else max(1, limit),
+        )
+        if manual_ids:
             logger.info(summary.start_log())
         try:
-            self._drain_search_queue(
+            processed = self._drain_search_queue(
                 queue=queue,
-                limit=max(1, limit),
+                limit=max(1, limit, len(manual_ids)),
                 progress_callback=progress_callback,
                 summary=summary,
             )
+            if manual_ids:
+                processed_subscribes = [item for item in subscribes if item.id in processed]
+                self._notify_manual_search(
+                    True,
+                    manual_ids[0] if len(manual_ids) == 1 else None,
+                    None if len(manual_ids) == 1 else manual_ids,
+                    subscribes,
+                    processed_subscribes,
+                )
         except Exception:
             summary.round_failed = True
             raise
         finally:
+            if not manual_ids and summary.processed:
+                summary.requested = summary.processed
+            if manual_ids or summary.processed or summary.round_failed:
                 logger.info(summary.finish_log())
 
     def cancel_search_batch(self, batch_id: str) -> bool:
@@ -750,15 +818,52 @@ class SubscribeSearchOwner(_SubscribeSearchQueueOwner):
         return cast(list[SubscriptionSnapshot], repository.list(self.get_states_for_search(state or "N")))
 
     @staticmethod
-    def _defer_recent_subscription(subscribe: SubscriptionSnapshot) -> bool:
-        """新增一分钟内保留 N 状态，为用户编辑筛选条件留出窗口。"""
-        if not subscribe.date:
-            return False
+    def _recent_subscription_retry_at(
+        subscribe: SubscriptionSnapshot,
+        source: str,
+    ) -> Optional[str]:
+        """新订阅保留一分钟编辑时间，并返回自动开始搜索的时间。"""
+        if source != "new" or not subscribe.date:
+            return None
+        try:
             subscribe_time = datetime.strptime(subscribe.date, "%Y-%m-%d %H:%M:%S")
-        if (datetime.now() - subscribe_time).total_seconds() >= 60:
-            return False
-        logger.debug(f"订阅标题：{subscribe.name} 新增小于1分钟，暂不搜索...")
-        return True
+        except ValueError:
+            logger.warning(f"订阅《{subscribe.name}》的添加时间无法识别，将直接开始搜索")
+            return None
+        remaining = _NEW_SUBSCRIPTION_EDIT_SECONDS - (
+            datetime.now() - subscribe_time
+        ).total_seconds()
+        if remaining <= 0:
+            return None
+        retry_seconds = max(1, int(remaining) + 1)
+        logger.debug(f"新订阅《{subscribe.name}》将在约 {retry_seconds} 秒后自动开始搜索")
+        return _retry_at_after(retry_seconds)
+
+    def _SubscribeChain__queue_new_subscription_search(
+        self,
+        subscribe_id: int,
+    ) -> Optional[str]:
+        """在订阅保存后安排自动搜索，定时扫描仍作为恢复保障。"""
+        queue: Optional[SubscriptionSearchRepository] = getattr(
+            self, "subscription_search_repository", None
+        )
+        if queue is None:
+            return None
+        subscribe = self.subscription_repository.get(subscribe_id)
+        if subscribe is None or subscribe.state != "N":
+            return None
+        available_at = self._recent_subscription_retry_at(subscribe, "new")
+        enqueued = queue.enqueue(
+            subscription_ids=(subscribe_id,),
+            source="new",
+            priority=50,
+            available_at_by_subscription={
+                subscribe_id: available_at or datetime.now(timezone.utc).isoformat(timespec="seconds")
+            },
+        )
+        if enqueued.created_count:
+            logger.info(f"已安排新订阅《{subscribe.name}》自动搜索，保存好设置后会自动开始")
+        return enqueued.active_batch_ids[0] if enqueued.active_batch_ids else None
 
     @staticmethod
     def _wait_before_scheduled_search(
@@ -771,9 +876,9 @@ class SubscribeSearchOwner(_SubscribeSearchQueueOwner):
         if sid or sids is not None or state not in {"R", "P"}:
             return
         sleep_time = random.randint(60, 300)
-        logger.debug(f"订阅搜索随机休眠 {sleep_time} 秒 ...")
+        logger.debug(f"为了避免连续访问站点，约 {sleep_time} 秒后继续搜索")
         if progress_callback:
-            progress_callback(text=f"订阅搜索随机休眠 {sleep_time} 秒后继续 ...")
+            progress_callback(text=f"为了避免连续访问站点，约 {sleep_time} 秒后继续搜索 ...")
         time.sleep(sleep_time)
 
     def _process_search_subscription(
@@ -791,7 +896,7 @@ class SubscribeSearchOwner(_SubscribeSearchQueueOwner):
         try:
             meta = build_subscribe_meta(subscribe)
         except ValueError:
-            logger.error(f"订阅 {subscribe.name} 类型错误：{subscribe.type}")
+            logger.error(f"订阅《{subscribe.name}》的媒体类型不受支持，暂时无法搜索")
             return subscribe
         mediainfo: MediaInfo = MediaChain().recognize_media(
             meta=meta,
@@ -965,7 +1070,7 @@ class SubscribeSearchOwner(_SubscribeSearchQueueOwner):
         subscribes: list[SubscriptionSnapshot],
         processed: list[SubscriptionSnapshot],
     ) -> None:
-        """为手动触发的搜索发布保持旧文案的系统消息。"""
+        """用清晰文案反馈手动搜索的完成或等待状态。"""
         if not manual:
             return
         if not subscribes:
@@ -974,7 +1079,7 @@ class SubscribeSearchOwner(_SubscribeSearchQueueOwner):
             message = (
                 f"{subscribes[0].name} 搜索完成！"
                 if processed
-                else f"{subscribes[0].name} 本轮未执行或未完成，将等待下一次正常调度。"
+                else f"{subscribes[0].name} 已安排搜索，系统会自动继续处理。"
             )
             self.messagehelper.put(message, title="订阅搜索", role="system")
         elif sids is not None:
@@ -984,6 +1089,6 @@ class SubscribeSearchOwner(_SubscribeSearchQueueOwner):
             message = (
                 "所有订阅搜索完成！"
                 if len(processed) == len(subscribes)
-                else "订阅搜索完成，部分订阅本轮未执行或未完成。"
+                else "订阅搜索已安排，暂未完成的项目会自动继续处理。"
             )
             self.messagehelper.put(message, title="订阅搜索", role="system")

app/db/adapters/subscription.py

@@ -769,9 +769,18 @@ class SessionSubscriptionRepository:
                 candidates.append(candidate)
         return candidates
 
-    async def list_search_ids(self, username: str, state: str) -> builtins.list[int]:
-        """异步读取用户指定状态下的订阅主键。"""
-        return [snapshot.id for snapshot in await self.async_list_by_username(username, state)]
+    async def list_search_ids(
+        self,
+        username: Optional[str],
+        state: str,
+    ) -> builtins.list[int]:
+        """异步读取用户或管理员全局范围内指定状态的订阅主键。"""
+        snapshots = (
+            await self.async_list_by_username(username, state)
+            if username is not None
+            else await self.async_list(state)
+        )
+        return [snapshot.id for snapshot in snapshots]
 
     async def stage_delete(self, subscribe_id: int) -> None:
         """异步暂存删除订阅。"""

app/db/adapters/subscriptionsearch.py

@@ -1,6 +1,7 @@
 """订阅搜索持久队列的 SQLAlchemy 适配器。"""
 
 from collections.abc import Callable, Mapping
+from datetime import datetime, timedelta, timezone
 from typing import Optional, TypeVar
 
 from sqlalchemy.orm import Session
@@ -19,6 +20,7 @@ from app.db.oper.subscriptionsearch import SubscriptionSearchOper
 from app.db.uow import SqlAlchemyUnitOfWork
 
 T = TypeVar("T")
+_BUSY_SITE_RETRY_SECONDS = 10
 
 
 def _batch(record: SubscriptionSearchBatch) -> SearchBatchSnapshot:
@@ -101,7 +103,7 @@ class TransactionalSubscriptionSearchRepository:
         """创建批次并返回 single-flight 合并计数。"""
         def operation(repository: SubscriptionSearchOper) -> SearchEnqueueResult:
             """在同一事务内创建批次和任务。"""
-            record, created, coalesced = repository.enqueue(
+            record, created, coalesced, active_batch_ids = repository.enqueue(
                 subscription_ids=subscription_ids,
                 source=source,
                 priority=priority,
@@ -111,6 +113,7 @@ class TransactionalSubscriptionSearchRepository:
                 batch=_batch(record),
                 created_count=created,
                 coalesced_count=coalesced,
+                active_batch_ids=active_batch_ids,
             )
 
         return self._write(operation)
@@ -183,13 +186,17 @@ class TransactionalSubscriptionSearchRepository:
         task_id: str,
         lease_token: str,
         available_at: str,
+        phase: str = "waiting_site_budget",
+        message: Optional[str] = None,
     ) -> bool:
-        """以站点预算的下一次可用时间重新排队任务。"""
+        """按指定时间和可见原因重新排队任务。"""
         return self._write(
             lambda repository: repository.defer_task(
                 task_id=task_id,
                 lease_token=lease_token,
                 available_at=available_at,
+                phase=phase,
+                message=message,
             )
         )
 
@@ -225,14 +232,32 @@ class TransactionalSubscriptionSearchRepository:
                 lease_seconds=lease_seconds,
             )
             retry_at = record.next_allowed_at
-            if not acquired and record.lease_token and record.lease_expires_at:
-                retry_at = max(retry_at, record.lease_expires_at)
+            wait_reason = None
+            now = datetime.now(timezone.utc)
+            cooldown_active = bool(
+                record.last_outcome not in {None, "success", "skipped"}
+                and record.next_allowed_at > now.isoformat(timespec="seconds")
+            )
+            lease_busy = bool(
+                record.lease_token
+                and record.lease_expires_at
+                and record.lease_expires_at > now.isoformat(timespec="seconds")
+            )
+            if not acquired and cooldown_active:
+                wait_reason = "cooldown"
+            elif not acquired and lease_busy:
+                wait_reason = "busy"
+                short_retry = (now + timedelta(seconds=_BUSY_SITE_RETRY_SECONDS)).isoformat(
+                    timespec="seconds"
+                )
+                retry_at = min(record.lease_expires_at, short_retry)
             return SiteBudgetClaim(
                 site_id=record.site_id,
                 acquired=acquired,
                 retry_at=retry_at,
                 consecutive_failures=record.consecutive_failures,
                 lease_token=record.lease_token if acquired else None,
+                wait_reason=wait_reason,
             )
 
         return self._write(operation)

app/db/adapters/subscriptionstatus.py

@@ -83,6 +83,7 @@ class SessionSubscriptionExecutionStatusRepository:
         """返回最近更新的批次，访问范围由应用服务依据任务校验。"""
         result = await self._session.execute(
             select(SubscriptionSearchBatch)
+            .where(SubscriptionSearchBatch.total_count > 0)
             .order_by(
                 SubscriptionSearchBatch.updated_at.desc(),
                 SubscriptionSearchBatch.id.desc(),

app/db/oper/subscribe.py

@@ -556,9 +556,13 @@ class SubscribeOper(DbOper):
             )
         return candidates
 
-    async def list_search_ids(self, username: str, state: str) -> List[int]:
-        """返回用户指定状态的订阅编号，不向应用用例暴露 ORM 列表。"""
-        subscribes = await self.async_list_by_username(username, state=state)
+    async def list_search_ids(self, username: Optional[str], state: str) -> List[int]:
+        """返回用户或管理员全局范围内指定状态的订阅编号。"""
+        subscribes = (
+            await self.async_list_by_username(username, state=state)
+            if username is not None
+            else await self.async_list(state=state)
+        )
         return [subscribe.id for subscribe in subscribes if subscribe.id]
 
     def get_by(

app/db/oper/subscriptionsearch.py

@@ -31,7 +31,7 @@ class SubscriptionSearchOper(DbOper):
         source: str,
         priority: int,
         available_at_by_subscription: Optional[Mapping[int, str]],
-    ) -> tuple[SubscriptionSearchBatch, int, int]:
+    ) -> tuple[SubscriptionSearchBatch, int, int, tuple[str, ...]]:
         """创建批次，并以活动键合并同一订阅的重叠搜索入口。"""
         if not isinstance(self._db, Session):
             raise RuntimeError("订阅搜索入队需要调用方提供同步 Session")
@@ -49,6 +49,7 @@ class SubscriptionSearchOper(DbOper):
         self._db.flush()
         created = 0
         coalesced = 0
+        active_batch_ids: list[str] = []
         for position, subscription_id in enumerate(dict.fromkeys(subscription_ids)):
             active_key = f"subscription:{subscription_id}"
             available_at = (
@@ -56,6 +57,11 @@ class SubscriptionSearchOper(DbOper):
                 if available_at_by_subscription
                 else now
             )
+            initial_phase = (
+                "scheduled"
+                if source == "new" and available_at > now
+                else "queued"
+            )
             task = SubscriptionSearchTask(
                 task_id=uuid4().hex,
                 batch_id=batch.batch_id,
@@ -65,7 +71,7 @@ class SubscriptionSearchOper(DbOper):
                 priority=priority,
                 position=position,
                 state="queued",
-                phase="queued",
+                phase=initial_phase,
                 available_at=available_at,
                 created_at=now,
                 updated_at=now,
@@ -77,15 +83,31 @@ class SubscriptionSearchOper(DbOper):
                 created += 1
             except IntegrityError:
                 coalesced += 1
+                promote_queued_task = and_(
+                    SubscriptionSearchTask.priority < priority,
+                    SubscriptionSearchTask.state == "queued",
+                )
                 execute_dml(
                     self._db,
                     update(SubscriptionSearchTask)
                     .where(SubscriptionSearchTask.active_key == active_key)
                     .values(
+                        source=case(
+                            (SubscriptionSearchTask.priority < priority, source),
+                            else_=SubscriptionSearchTask.source,
+                        ),
                         priority=case(
                             (SubscriptionSearchTask.priority < priority, priority),
                             else_=SubscriptionSearchTask.priority,
                         ),
+                        phase=case(
+                            (promote_queued_task, "queued"),
+                            else_=SubscriptionSearchTask.phase,
+                        ),
+                        last_error=case(
+                            (promote_queued_task, None),
+                            else_=SubscriptionSearchTask.last_error,
+                        ),
                         available_at=case(
                             (
                                 or_(
@@ -100,11 +122,20 @@ class SubscriptionSearchOper(DbOper):
                     ),
                     execution_options={"synchronize_session": False},
                 )
+                active_task = self._db.execute(
+                    select(SubscriptionSearchTask).where(
+                        SubscriptionSearchTask.active_key == active_key
+                    )
+                ).scalars().first()
+                if active_task is not None:
+                    active_batch_ids.append(active_task.batch_id)
         batch.total_count = created
         if created == 0:
             batch.state = "completed"
             batch.finished_at = now
-        return batch, created, coalesced
+        else:
+            active_batch_ids.insert(0, batch.batch_id)
+        return batch, created, coalesced, tuple(dict.fromkeys(active_batch_ids))
 
     def claim_next(self, *, owner: str, lease_seconds: int) -> Optional[SubscriptionSearchTask]:
         """使用 CAS 认领最高优先级任务，过期 running 任务可被恢复。"""
@@ -139,6 +170,7 @@ class SubscriptionSearchOper(DbOper):
                 )
                 .order_by(
                     case(
+                        (SubscriptionSearchTask.priority >= 100, 2),
                         (SubscriptionSearchTask.created_at <= fairness_before, 1),
                         else_=0,
                     ).desc(),
@@ -369,8 +401,10 @@ class SubscriptionSearchOper(DbOper):
         task_id: str,
         lease_token: str,
         available_at: str,
+        phase: str,
+        message: Optional[str],
     ) -> bool:
-        """释放当前租约并在站点预算时间到达后恢复同一任务。"""
+        """释放当前租约并在指定时间后按可见原因恢复同一任务。"""
         if not isinstance(self._db, Session):
             raise RuntimeError("订阅搜索延后需要调用方提供同步 Session")
         task = self._db.execute(
@@ -400,7 +434,7 @@ class SubscriptionSearchOper(DbOper):
             )
             .values(
                 state="queued",
-                phase="waiting_site_budget",
+                phase=phase,
                 current_site_id=None,
                 lease_owner=None,
                 lease_token=None,
@@ -408,7 +442,7 @@ class SubscriptionSearchOper(DbOper):
                 available_at=available_at,
                 updated_at=now,
                 finished_at=None,
-                last_error=None,
+                last_error=message,
             ),
             execution_options={"synchronize_session": False},
         )

app/locales/en-US.json

@@ -137,6 +137,17 @@
     "当前管理用户缺少可审计身份": "The current administrator does not have an auditable identity",
     "人工复核任务不存在": "The manual review task does not exist",
     "任务添加失败": "Failed to add task",
+    "没有需要搜索的订阅": "No subscriptions need searching",
+    "这个订阅已经在搜索中，请稍候": "This subscription is already being searched. Please wait",
+    "已安排搜索，很快开始": "Search scheduled and will start soon",
+    "当前没有可检查的订阅资源": "No subscription resources are currently available to check",
+    "正在整理订阅资源 ...": "Organizing subscription resources ...",
+    "订阅资源检查完成": "Subscription resource check complete",
+    "订阅资源检查已停止，部分订阅这次未检查": "Subscription resource check stopped; some subscriptions were not checked this time",
+    "订阅资源检查结束，部分订阅没有完成": "Subscription resource check finished; some subscriptions did not complete",
+    "订阅资源检查完成，部分订阅这次未检查": "Subscription resource check complete; some subscriptions were not checked this time",
+    "订阅搜索正在处理中，本次不再重复开始": "A subscription search is already in progress. This run will not start again",
+    "订阅资源检查正在进行，本次不再重复开始": "A subscription resource check is already in progress. This run will not start again",
     "无法识别媒体信息": "Unable to recognize media information",
     "未识别到媒体信息": "Unable to recognize media information",
     "未识别到音乐信息": "Unable to recognize music information",
@@ -382,8 +393,8 @@
     "微信 ClawBot 通知未启用或配置尚未保存，请先保存并启用当前渠道": "WeChat ClawBot notification is not enabled or the configuration has not been saved. Please save and enable this channel first",
     "请输入至少一个有效的站点 ID": "Enter at least one valid site ID",
     "所有订阅搜索完成": "All subscription searches are complete",
-    "订阅搜索批次不存在": "The subscription search batch does not exist",
-    "订阅搜索批次已结束或无法取消": "The subscription search batch has ended or cannot be cancelled",
+    "没有找到这次搜索，请刷新后重试": "This search could not be found. Refresh and try again",
+    "这次搜索已经结束，暂时无法停止": "This search has already ended and cannot be stopped",
     "订阅搜索锁等待超时，已跳过本轮": "Subscription search lock timed out, this round was skipped",
     "订阅匹配锁等待超时，已跳过本轮": "Subscription matching lock timed out, this round was skipped",
     "请输入订阅 ID，多个 ID 用空格分隔，或输入 all": "Enter subscription IDs separated by spaces, or enter all",
@@ -765,6 +776,18 @@
       "source": "已完成 {count} 个订阅搜索",
       "target": "Completed searches for {count} subscriptions"
     },
+    {
+      "source": "{count} 个订阅已经在搜索中，无需重复提交",
+      "target": "{count} subscriptions are already being searched. No need to submit again"
+    },
+    {
+      "source": "已安排 {queued} 个订阅搜索，另有 {ongoing} 个正在处理中",
+      "target": "Scheduled searches for {queued} subscriptions; {ongoing} are already being processed"
+    },
+    {
+      "source": "已安排 {count} 个订阅搜索，系统会依次处理",
+      "target": "Scheduled searches for {count} subscriptions. They will be processed in order"
+    },
     {
       "source": "请输入至少一个有效的订阅 ID",
       "target": "Enter at least one valid subscription ID"
@@ -1421,6 +1444,10 @@
       "source": "订阅搜索（{index}/{total}）处理完成",
       "target": "Subscription search ({index}/{total}) completed"
     },
+    {
+      "source": "为了避免连续访问站点，约 {seconds} 秒后继续搜索 ...",
+      "target": "To avoid repeated site requests, search will continue in about {seconds} seconds ..."
+    },
     {
       "source": "开始刷新订阅，共 {count} 个订阅 ...",
       "target": "Starting subscription refresh, {count} subscriptions ..."
@@ -1433,6 +1460,18 @@
       "source": "正在匹配订阅（{index}/{total}）{name} ...",
       "target": "Matching subscription ({index}/{total}) {name} ..."
     },
+    {
+      "source": "资源整理完成，开始检查 {count} 个订阅 ...",
+      "target": "Resource organization complete. Checking {count} subscriptions ..."
+    },
+    {
+      "source": "正在检查订阅（{index}/{total}）{name} ...",
+      "target": "Checking subscription ({index}/{total}) {name} ..."
+    },
+    {
+      "source": "已检查订阅（{index}/{total}）",
+      "target": "Checked subscription ({index}/{total})"
+    },
     {
       "source": "开始更新订阅元数据，共 {count} 个订阅 ...",
       "target": "Starting subscription metadata update, {count} subscriptions ..."

app/locales/zh-TW.json

@@ -129,6 +129,17 @@
     "当前管理用户缺少可审计身份": "目前管理使用者缺少可稽核身分",
     "人工复核任务不存在": "人工複核任務不存在",
     "任务添加失败": "任務新增失敗",
+    "没有需要搜索的订阅": "沒有需要搜尋的訂閱",
+    "这个订阅已经在搜索中，请稍候": "這個訂閱已在搜尋中，請稍候",
+    "已安排搜索，很快开始": "已安排搜尋，很快開始",
+    "当前没有可检查的订阅资源": "目前沒有可檢查的訂閱資源",
+    "正在整理订阅资源 ...": "正在整理訂閱資源 ...",
+    "订阅资源检查完成": "訂閱資源檢查完成",
+    "订阅资源检查已停止，部分订阅这次未检查": "訂閱資源檢查已停止，部分訂閱這次未檢查",
+    "订阅资源检查结束，部分订阅没有完成": "訂閱資源檢查結束，部分訂閱沒有完成",
+    "订阅资源检查完成，部分订阅这次未检查": "訂閱資源檢查完成，部分訂閱這次未檢查",
+    "订阅搜索正在处理中，本次不再重复开始": "訂閱搜尋正在處理中，本次不再重複開始",
+    "订阅资源检查正在进行，本次不再重复开始": "訂閱資源檢查正在進行，本次不再重複開始",
     "无法识别媒体信息": "無法識別媒體資訊",
     "未识别到媒体信息": "未識別到媒體資訊",
     "未识别到音乐信息": "未識別到音樂資訊",
@@ -374,8 +385,8 @@
     "微信 ClawBot 通知未启用或配置尚未保存，请先保存并启用当前渠道": "微信 ClawBot 通知未啟用或設定尚未儲存，請先儲存並啟用目前渠道",
     "请输入至少一个有效的站点 ID": "請輸入至少一個有效的站點 ID",
     "所有订阅搜索完成": "所有訂閱搜尋完成",
-    "订阅搜索批次不存在": "訂閱搜尋批次不存在",
-    "订阅搜索批次已结束或无法取消": "訂閱搜尋批次已結束或無法取消",
+    "没有找到这次搜索，请刷新后重试": "找不到這次搜尋，請重新整理後再試",
+    "这次搜索已经结束，暂时无法停止": "這次搜尋已經結束，暫時無法停止",
     "请输入订阅 ID，多个 ID 用空格分隔，或输入 all": "請輸入訂閱 ID，多個 ID 以空格分隔，或輸入 all",
     "请输入至少一个有效的订阅 ID": "請輸入至少一個有效的訂閱 ID",
     "格式错误，请输入：cookie <id> <username> <password> [2fa_code/secret]": "格式錯誤，請輸入：cookie <id> <username> <password> [2fa_code/secret]",
@@ -751,6 +762,18 @@
       "source": "已完成 {count} 个订阅搜索",
       "target": "已完成 {count} 個訂閱搜尋"
     },
+    {
+      "source": "{count} 个订阅已经在搜索中，无需重复提交",
+      "target": "{count} 個訂閱已在搜尋中，無需重複提交"
+    },
+    {
+      "source": "已安排 {queued} 个订阅搜索，另有 {ongoing} 个正在处理中",
+      "target": "已安排 {queued} 個訂閱搜尋，另有 {ongoing} 個正在處理中"
+    },
+    {
+      "source": "已安排 {count} 个订阅搜索，系统会依次处理",
+      "target": "已安排 {count} 個訂閱搜尋，系統會依序處理"
+    },
     {
       "source": "请输入至少一个有效的订阅 ID",
       "target": "請輸入至少一個有效的訂閱 ID"
@@ -1407,6 +1430,10 @@
       "source": "订阅搜索（{index}/{total}）处理完成",
       "target": "訂閱搜尋（{index}/{total}）處理完成"
     },
+    {
+      "source": "为了避免连续访问站点，约 {seconds} 秒后继续搜索 ...",
+      "target": "為了避免連續存取站點，約 {seconds} 秒後繼續搜尋 ..."
+    },
     {
       "source": "开始刷新订阅，共 {count} 个订阅 ...",
       "target": "開始重新整理訂閱，共 {count} 個訂閱 ..."
@@ -1419,6 +1446,18 @@
       "source": "正在匹配订阅（{index}/{total}）{name} ...",
       "target": "正在匹配訂閱（{index}/{total}）{name} ..."
     },
+    {
+      "source": "资源整理完成，开始检查 {count} 个订阅 ...",
+      "target": "資源整理完成，開始檢查 {count} 個訂閱 ..."
+    },
+    {
+      "source": "正在检查订阅（{index}/{total}）{name} ...",
+      "target": "正在檢查訂閱（{index}/{total}）{name} ..."
+    },
+    {
+      "source": "已检查订阅（{index}/{total}）",
+      "target": "已檢查訂閱（{index}/{total}）"
+    },
     {
       "source": "开始更新订阅元数据，共 {count} 个订阅 ...",
       "target": "開始更新訂閱元資料，共 {count} 個訂閱 ..."

app/scheduler/catalog.py

@@ -154,8 +154,8 @@ class SchedulerCatalogOwner(_SchedulerOwnerBase):
             "interval",
             id="subscribe_search_queue",
             name="恢复订阅搜索队列",
-            minutes=1,
-            next_run_time=datetime.now(pytz.timezone(config.timezone)) + timedelta(seconds=10),
+            seconds=10,
+            next_run_time=datetime.now(pytz.timezone(config.timezone)) + timedelta(seconds=5),
             kwargs={"job_id": "subscribe_search_queue"},
         )
 

app/schemas/exports.py

@@ -463,6 +463,7 @@ SCHEMA_EXPORTS = {
     'SubscriptionBatchStatus': ('app.schemas.subscribe', 'SubscriptionBatchStatus'),
     'SubscriptionExecutionStatus': ('app.schemas.subscribe', 'SubscriptionExecutionStatus'),
     'SubscriptionMessage': ('app.schemas.message', 'SubscriptionMessage'),
+    'SubscriptionSearchSubmission': ('app.schemas.subscribe', 'SubscriptionSearchSubmission'),
     'SubtitleDownloadData': ('app.schemas.download', 'SubtitleDownloadData'),
     'SubtitleInfo': ('app.schemas.search', 'SubtitleInfo'),
     'SystemEnvironmentUpdateData': ('app.schemas.system', 'SystemEnvironmentUpdateData'),

app/schemas/subscribe.py

@@ -55,6 +55,7 @@ class SubscriptionExecutionStatus(BaseModel):  # type: ignore[misc]
     batch_id: Optional[str] = None
     task_id: Optional[str] = None
     current_site_id: Optional[int] = None
+    next_run_at: Optional[str] = None
     error: Optional[str] = None
     can_cancel: bool = False
 
@@ -84,6 +85,17 @@ class SubscriptionBatchStatus(BaseModel):  # type: ignore[misc]
     model_config = ConfigDict(from_attributes=True)
 
 
+class SubscriptionSearchSubmission(BaseModel):  # type: ignore[misc]
+    """手工订阅搜索已安排后的轻量跟踪信息。"""
+
+    batch_id: Optional[str] = None
+    batch_ids: List[str] = Field(default_factory=list)
+    target_count: int = 0
+    queued_count: int = 0
+    ongoing_count: int = 0
+    single: bool = False
+
+
 class Subscribe(OptionalMediaIdentityMixin, BaseModel):
     """订阅输入与响应模型，媒体身份必须为空对或完整有效对。"""
 

scripts/validation/subscription_governance_scale.py

@@ -799,7 +799,7 @@ def _run_match_execution_case(case: ScaleCase) -> dict[str, Any]:
 
         @staticmethod
         def info(message: Any, *_args: Any, **_kwargs: Any) -> None:
-            """保存结构化批次摘要。"""
+            """保存开始和结束摘要。"""
             info_logs.append(str(message))
 
     class _ScaleMediaChain:
@@ -958,8 +958,10 @@ def _run_match_execution_case(case: ScaleCase) -> dict[str, Any]:
         "info_log_count": len(info_logs),
         "info_log_bounded": bool(
             len(info_logs) == 2
-            and info_logs[0].startswith("订阅治理轮次开始: operation=match ")
-            and info_logs[1].startswith("订阅治理轮次结束: operation=match ")
+            and info_logs[0].startswith("开始检查订阅资源，共 ")
+            and info_logs[1].startswith("订阅资源检查结束：")
+            and "operation=" not in "\n".join(info_logs)
+            and "run_id=" not in "\n".join(info_logs)
         ),
         "first_settlement_local_ms": round(settlement_ms[0], 3),
         "subscription_settlement_local_p50_ms": _percentile(settlement_ms, 50),

tests/test_api_background_task_registry.py

@@ -72,6 +72,34 @@ class _RunningTaskRegistry(TaskRegistry):
         )
 
 
+class _CompletedSearchTaskRegistry(_TaskRegistry):
+    """返回已完成结果，验证手工搜索先保存安排再启动后台处理。"""
+
+    def create_sync(self, function, *args, owner: str, **kwargs) -> asyncio.Future:
+        """记录同步任务，并为测试提供可等待的安排结果。"""
+        super().create_sync(function, *args, owner=owner, **kwargs)
+        future = asyncio.get_running_loop().create_future()
+        result = None
+        if owner == "api.subscribe.search.enqueue":
+            result = SimpleNamespace(
+                active_batch_ids=("batch-1", "batch-2"),
+                created_count=2,
+                coalesced_count=0,
+            )
+        future.set_result(result)
+        return future
+
+
+class _SubscriptionSearchTargets:
+    """为手工搜索命令提供当前用户可访问的订阅编号。"""
+
+    async def list_search_ids(self, username, state) -> list[int]:
+        """返回稳定目标，并校验超级用户读取全部运行中订阅。"""
+        assert username is None
+        assert state == "R"
+        return [11, 12]
+
+
 class _ProtocolManager:
     """提供兼容协议流结束时需要的最小 AgentManager 接口。"""
 
@@ -206,10 +234,14 @@ def test_seerr_subscribe_uses_task_registry(monkeypatch) -> None:
 
 def test_manual_subscription_search_uses_task_registry() -> None:
     """手工订阅搜索命令应以稳定 owner 提交顺序搜索批次。"""
-    registry = _TaskRegistry()
-    repository = object()
+    registry = _CompletedSearchTaskRegistry()
+    repository = _SubscriptionSearchTargets()
+    search_repository = SimpleNamespace(enqueue=lambda **_kwargs: None)
     runtime = SimpleNamespace(
-        subscription=SimpleNamespace(repository=lambda _db: repository)
+        subscription=SimpleNamespace(
+            repository=lambda _db: repository,
+            search_repository=search_repository,
+        )
     )
     command = subscription_dependencies.get_search_subscriptions_command(
         task_registry=registry,
@@ -221,12 +253,24 @@ def test_manual_subscription_search_uses_task_registry() -> None:
         command.execute(SubscribeSearchActor(username="admin", is_superuser=True))
     )
 
-    function, args, kwargs, owner = registry.calls[0]
-    assert found is True
-    assert function is subscription_dependencies._start_subscription_search_batch
-    assert args == (None, "R")
-    assert kwargs == {}
-    assert owner == "api.subscribe.search"
+    enqueue_function, enqueue_args, enqueue_kwargs, enqueue_owner = registry.calls[0]
+    run_function, run_args, run_kwargs, run_owner = registry.calls[1]
+    assert found is not None
+    assert found.batch_ids == ("batch-1", "batch-2")
+    assert found.queued_count == 2
+    assert found.ongoing_count == 0
+    assert enqueue_function is search_repository.enqueue
+    assert enqueue_args == ()
+    assert enqueue_kwargs == {
+        "subscription_ids": (11, 12),
+        "source": "manual",
+        "priority": 100,
+    }
+    assert enqueue_owner == "api.subscribe.search.enqueue"
+    assert run_function is subscription_dependencies._resume_submitted_subscription_search
+    assert run_args == ((11, 12),)
+    assert run_kwargs == {}
+    assert run_owner == "api.subscribe.search.run"
 
 
 def test_history_ai_redo_uses_task_registry() -> None:

tests/test_subscribe_endpoint.py

@@ -653,8 +653,19 @@ class TestSubscribeEndpoint:
         普通用户批量搜索把用户身份交给应用命令。
         """
         from app.api.endpoints.subscribe import search_subscribes
+        from app.application.subscription.search import SubscriptionSearchSubmission
 
-        command = SimpleNamespace(execute=AsyncMock(return_value=True))
+        command = SimpleNamespace(
+            execute=AsyncMock(
+                return_value=SubscriptionSearchSubmission(
+                    batch_ids=("batch-1",),
+                    target_count=2,
+                    queued_count=1,
+                    ongoing_count=1,
+                    single=False,
+                )
+            )
+        )
         response = asyncio.run(
             search_subscribes(
                 command=command,
@@ -663,6 +674,8 @@ class TestSubscribeEndpoint:
         )
 
         assert response.success
+        assert response.message == "已安排 1 个订阅搜索，另有 1 个正在处理中"
+        assert response.data.queued_count == 1
         command.execute.assert_awaited_once()
         actor = command.execute.await_args.args[0]
         assert actor.username == "alice"

tests/test_subscribe_search_command.py

@@ -1,3 +1,4 @@
+import asyncio
 from types import SimpleNamespace
 from unittest.mock import Mock
 
@@ -8,6 +9,7 @@ from app.application.subscription.delete import SubscribeDeletionCandidate
 from app.application.subscription.search import (
     SearchSubscriptionsCommand,
     SubscribeSearchActor,
+    SubscriptionSearchSubmission,
 )
 from app.runtime.tasks import TaskRegistry
 
@@ -19,15 +21,15 @@ class _Repository:
         """保存预设单条候选和批量编号。"""
         self.candidate = candidate
         self.subscribe_ids = subscribe_ids or []
+        self.list_calls = []
 
     async def get_candidate(self, _subscribe_id):
         """返回预设订阅候选。"""
         return self.candidate
 
     async def list_search_ids(self, username, state):
-        """校验普通用户搜索状态并返回预设编号。"""
-        assert username == "alice"
-        assert state == "R"
+        """记录访问范围并返回预设编号。"""
+        self.list_calls.append((username, state))
         return self.subscribe_ids
 
 
@@ -40,85 +42,117 @@ def _candidate(username):
     )
 
 
-@pytest.mark.asyncio
-async def test_superuser_search_all_uses_single_global_scheduler_request():
-    """管理员搜索全部订阅时保持一次 state=R 的全局调度语义。"""
-    scheduled = []
-    command = SearchSubscriptionsCommand(
-        repository=_Repository(),
-        schedule_search=lambda sid, state: scheduled.append((sid, state)),
+def _submitter(calls):
+    """构造记录目标并返回稳定安排结果的异步提交器。"""
+    async def submit(subscribe_ids, single):
+        """记录一次轻量入队请求。"""
+        calls.append((subscribe_ids, single))
+        return SubscriptionSearchSubmission(
+            batch_ids=("batch-1",),
+            target_count=len(subscribe_ids),
+            queued_count=len(subscribe_ids),
+            ongoing_count=0,
+            single=single,
         )
 
-    assert await command.execute(
+    return submit
+
+
+@pytest.mark.asyncio
+async def test_superuser_search_all_queues_every_active_subscription():
+    """管理员搜索全部订阅时读取全局活动订阅并一次提交。"""
+    submitted = []
+    repository = _Repository(subscribe_ids=[2, 5])
+    command = SearchSubscriptionsCommand(
+        repository=repository,
+        submit_search=_submitter(submitted),
+    )
+
+    result = await command.execute(
         SubscribeSearchActor(username="admin", is_superuser=True)
-    ) is True
-    assert scheduled == [(None, "R")]
+    )
+
+    assert result is not None
+    assert result.target_count == 2
+    assert repository.list_calls == [(None, "R")]
+    assert submitted == [((2, 5), False)]
 
 
 @pytest.mark.asyncio
 async def test_regular_user_search_all_schedules_only_owned_subscriptions():
     """普通用户搜索全部时把归属订阅合并为一次后台批次。"""
-    scheduled = []
+    submitted = []
+    repository = _Repository(subscribe_ids=[2, 5])
     command = SearchSubscriptionsCommand(
-        repository=_Repository(subscribe_ids=[2, 5]),
-        schedule_search=lambda sid, state: scheduled.append((sid, state)),
+        repository=repository,
+        submit_search=_submitter(submitted),
     )
 
-    assert await command.execute(
+    result = await command.execute(
         SubscribeSearchActor(username="alice", is_superuser=False)
-    ) is True
-    assert scheduled == [((2, 5), None)]
+    )
+
+    assert result is not None
+    assert repository.list_calls == [("alice", "R")]
+    assert submitted == [((2, 5), False)]
 
 
 @pytest.mark.asyncio
 async def test_regular_user_search_all_with_no_targets_does_not_schedule():
     """普通用户没有可搜索订阅时不创建空后台任务。"""
-    scheduled = []
+    submitted = []
     command = SearchSubscriptionsCommand(
         repository=_Repository(),
-        schedule_search=lambda ids, state: scheduled.append((ids, state)),
+        submit_search=_submitter(submitted),
     )
 
-    assert await command.execute(
+    result = await command.execute(
         SubscribeSearchActor(username="alice", is_superuser=False)
-    ) is True
-    assert scheduled == []
+    )
+
+    assert result is not None
+    assert result.target_count == 0
+    assert result.batch_ids == ()
+    assert submitted == []
 
 
 @pytest.mark.asyncio
 async def test_targeted_search_rejects_missing_or_other_users_subscription():
     """单条搜索不得泄漏订阅是否属于其他普通用户。"""
-    scheduled = []
+    submitted = []
     command = SearchSubscriptionsCommand(
         repository=_Repository(candidate=_candidate("bob")),
-        schedule_search=lambda sid, state: scheduled.append((sid, state)),
+        submit_search=_submitter(submitted),
     )
 
     assert await command.execute(
         SubscribeSearchActor(username="alice", is_superuser=False),
         subscribe_id=7,
-    ) is False
-    assert scheduled == []
+    ) is None
+    assert submitted == []
 
 
 @pytest.mark.asyncio
 async def test_targeted_search_schedules_accessible_subscription():
     """归属用户搜索单条订阅时提交历史兼容参数。"""
-    scheduled = []
+    submitted = []
     command = SearchSubscriptionsCommand(
         repository=_Repository(candidate=_candidate("alice")),
-        schedule_search=lambda sid, state: scheduled.append((sid, state)),
+        submit_search=_submitter(submitted),
     )
 
-    assert await command.execute(
+    result = await command.execute(
         SubscribeSearchActor(username="alice", is_superuser=False),
         subscribe_id=7,
-    ) is True
-    assert scheduled == [((7,), None)]
+    )
+
+    assert result is not None
+    assert result.single is True
+    assert submitted == [((7,), True)]
 
 
-def test_subscription_search_batch_uses_one_scheduler_generation(monkeypatch):
-    """一个后台批次只占用一次调度任务运行权。"""
+def test_submitted_search_wakes_short_cycle_queue(monkeypatch):
+    """手工搜索入队后立即唤醒短周期恢复任务。"""
     calls = []
     monkeypatch.setattr(
         subscription_dependencies,
@@ -126,24 +160,46 @@ def test_subscription_search_batch_uses_one_scheduler_generation(monkeypatch):
         lambda job_id, **kwargs: calls.append((job_id, kwargs)),
     )
 
-    subscription_dependencies._start_subscription_search_batch((2, 5), None)
+    subscription_dependencies._resume_submitted_subscription_search((2, 5))
 
     assert calls == [
         (
-            "subscribe_search",
-            {"sids": (2, 5), "state": None, "manual": True},
+            "subscribe_search_queue",
+            {"limit": 2, "manual_sids": (2, 5)},
         ),
     ]
 
 
 @pytest.mark.asyncio
-async def test_search_dependency_registers_one_owned_background_batch():
-    """请求适配器只登记一个具名后台批次。"""
+async def test_search_dependency_persists_before_waking_background_worker():
+    """请求适配器先等待持久入队完成，再登记后台恢复任务。"""
     registry = TaskRegistry()
-    registry.create_sync = Mock()
+    calls = []
+
+    def create_sync(function, *args, owner, **kwargs):
+        """同步执行测试函数，并返回已经完成的 Future。"""
+        calls.append((function, args, owner, kwargs))
+        future = asyncio.get_running_loop().create_future()
+        result = function(*args, **kwargs) if owner.endswith("enqueue") else None
+        future.set_result(result)
+        return future
+
+    registry.create_sync = Mock(side_effect=create_sync)
     repository = _Repository(subscribe_ids=[2, 5])
+    search_repository = SimpleNamespace(
+        enqueue=Mock(
+            return_value=SimpleNamespace(
+                active_batch_ids=("batch-new", "batch-existing"),
+                created_count=1,
+                coalesced_count=1,
+            )
+        )
+    )
     runtime = SimpleNamespace(
-        subscription=SimpleNamespace(repository=lambda _db: repository),
+        subscription=SimpleNamespace(
+            repository=lambda _db: repository,
+            search_repository=search_repository,
+        ),
     )
     command = subscription_dependencies.get_search_subscriptions_command(
         task_registry=registry,
@@ -151,12 +207,23 @@ async def test_search_dependency_registers_one_owned_background_batch():
         runtime=runtime,
     )
 
-    assert await command.execute(
+    result = await command.execute(
         SubscribeSearchActor(username="alice", is_superuser=False)
-    ) is True
-    registry.create_sync.assert_called_once_with(
-        subscription_dependencies._start_subscription_search_batch,
-        (2, 5),
-        None,
-        owner="api.subscribe.search",
+    )
+
+    assert result is not None
+    assert result.batch_ids == ("batch-new", "batch-existing")
+    assert result.queued_count == 1
+    assert result.ongoing_count == 1
+    assert calls[0] == (
+        search_repository.enqueue,
+        (),
+        "api.subscribe.search.enqueue",
+        {"subscription_ids": (2, 5), "source": "manual", "priority": 100},
+    )
+    assert calls[1] == (
+        subscription_dependencies._resume_submitted_subscription_search,
+        ((2, 5),),
+        "api.subscribe.search.run",
+        {},
     )

tests/test_subscribe_search_state.py

@@ -175,7 +175,7 @@ def test_subscribe_search_aborts_when_lock_times_out(monkeypatch) -> None:
     subscribe_oper.assert_not_called()
     progress.assert_called_once_with(
         value=100,
-        text="订阅搜索锁等待超时，已跳过本轮",
+        text="订阅搜索正在处理中，本次不再重复开始",
     )
 
 
@@ -265,7 +265,7 @@ def test_inline_search_conflict_does_not_report_false_completion(monkeypatch) ->
 
     assert progress.call_args.kwargs == {
         "value": 100,
-        "text": "订阅搜索结束，部分订阅本轮未执行或未完成",
+        "text": "搜索结束，部分订阅这次没有完成",
         "data": {"total": 1, "finished": 0},
     }
     assert chain._subscription_execution_admission.release(match_lease) is True
@@ -279,4 +279,4 @@ def test_subscribe_match_aborts_when_lock_times_out(monkeypatch) -> None:
     chain = object.__new__(SubscribeChain)
     chain.match({"example.org": []}, progress_callback=progress)
 
-    progress.assert_any_call(value=100, text="订阅匹配锁等待超时，已跳过本轮")
+    progress.assert_any_call(value=100, text="订阅资源检查正在进行，本次不再重复开始")

tests/test_subscription_execution_status.py

@@ -14,13 +14,15 @@ def _task(
     phase: str = "searching",
     updated_at: str = "2026-09-01T01:00:00+00:00",
     batch_id: str = "batch-1",
+    source: str = "manual",
+    available_at: str | None = None,
 ) -> SearchTaskSnapshot:
     """构造最小搜索任务快照。"""
     return SearchTaskSnapshot(
         task_id=f"task-{subscription_id}",
         batch_id=batch_id,
         subscription_id=subscription_id,
-        source="manual",
+        source=source,
         priority=100,
         position=subscription_id,
         state=state,
@@ -30,8 +32,15 @@ def _task(
         lease_token="lease" if state == "running" else None,
         created_at="2026-09-01T00:00:00+00:00",
         updated_at=updated_at,
+        available_at=available_at,
         current_site_id=9 if phase == "waiting_site_budget" else None,
-        last_error=" provider\n timeout " if state == "failed" else None,
+        last_error=(
+            " provider\n timeout "
+            if state == "failed"
+            else "站点暂时忙，系统会自动继续搜索"
+            if state == "queued" and phase == "waiting_site_budget"
+            else None
+        ),
     )
 
 
@@ -85,18 +94,44 @@ def test_execution_status_exposes_site_wait_and_cancel_capability():
 
 
 def test_execution_status_exposes_queued_site_wait_without_error():
-    """重新入队的站点预算冲突应显示等待状态而不是失败。"""
+    """重新入队的站点繁忙应显示等待状态、说明和继续时间。"""
     repository = _Repository()
-    repository.tasks[2] = _task(2, state="queued", phase="waiting_site_budget")
+    retry_at = "2026-09-01T01:00:10+00:00"
+    repository.tasks[2] = _task(
+        2,
+        state="queued",
+        phase="waiting_site_budget",
+        available_at=retry_at,
+    )
 
     statuses = asyncio.run(SubscriptionExecutionStatusService(repository).for_subscriptions((2,)))
 
     assert statuses[2].state == "waiting_site_budget"
     assert statuses[2].phase == "waiting_site_budget"
-    assert statuses[2].error is None
+    assert statuses[2].error == "站点暂时忙，系统会自动继续搜索"
+    assert statuses[2].next_run_at == retry_at
     assert statuses[2].can_cancel is True
 
 
+def test_execution_status_exposes_scheduled_new_search_without_failure():
+    """新订阅编辑等待期应显示为已安排，而不是跳过或失败。"""
+    repository = _Repository()
+    retry_at = "2026-09-01T01:01:00+00:00"
+    repository.tasks[4] = _task(
+        4,
+        state="queued",
+        phase="scheduled",
+        source="new",
+        available_at=retry_at,
+    )
+
+    statuses = asyncio.run(SubscriptionExecutionStatusService(repository).for_subscriptions((4,)))
+
+    assert statuses[4].state == "scheduled"
+    assert statuses[4].next_run_at == retry_at
+    assert statuses[4].error is None
+
+
 def test_failed_search_exposes_safe_error():
     """搜索失败文本必须压平且不暴露内部错误细节。"""
     repository = _Repository()

tests/test_subscription_match_admission.py

@@ -169,7 +169,7 @@ def test_match_skips_subscription_paused_after_admission() -> None:
         "skipped": 1,
         "failed": 0,
     }
-    assert progress.call_args.kwargs["text"] == "订阅资源匹配完成，部分订阅跳过"
+    assert progress.call_args.kwargs["text"] == "订阅资源检查完成，部分订阅这次未检查"
     _assert_channel_and_subscription_released(chain, (listed.id,))
 
 
@@ -200,7 +200,7 @@ def test_match_progress_reports_completed_skipped_and_failed_counts() -> None:
         "skipped": 1,
         "failed": 1,
     }
-    assert progress.call_args.kwargs["text"] == "订阅资源匹配完成，部分订阅失败"
+    assert progress.call_args.kwargs["text"] == "订阅资源检查结束，部分订阅没有完成"
     assert chain._subscription_execution_admission.release(skipped_lease) is True
     _assert_channel_and_subscription_released(chain, (completed.id, failed.id))
 
@@ -220,7 +220,7 @@ def test_match_stop_does_not_report_unvisited_subscriptions_as_completed(monkeyp
 
     assert progress.call_args.kwargs == {
         "value": 100,
-        "text": "订阅资源匹配已停止，部分订阅未执行",
+        "text": "订阅资源检查已停止，部分订阅这次未检查",
         "data": {
             "total": 2,
             "finished": 0,
@@ -257,17 +257,12 @@ def test_match_logs_one_bounded_start_and_finish_summary(monkeypatch) -> None:
     chain.match({"site-a.example": [object()], "site-b.example": [object(), object()]})
 
     assert len(info_logs) == 2
-    assert info_logs[0].startswith("订阅治理轮次开始: operation=match ")
-    assert "subscriptions=3" in info_logs[0]
-    assert "sites=2" in info_logs[0]
-    assert "candidates=3" in info_logs[0]
-    assert info_logs[1].startswith("订阅治理轮次结束: operation=match ")
-    assert "state=skipped" in info_logs[1]
-    assert "task_completed=1" in info_logs[1]
-    assert "task_skipped=2" in info_logs[1]
-    assert "admission_conflicts=1" in info_logs[1]
-    assert "cancelled=1" in info_logs[1]
-    assert "ttl_timeouts=0" in info_logs[1]
+    assert info_logs[0] == "开始检查订阅资源，共 3 个订阅、3 个资源，来自 2 个站点。"
+    assert info_logs[1].startswith("订阅资源检查结束：部分订阅这次未检查。")
+    assert "本次检查 3/3 个订阅，完成 1 个，这次未检查 2 个，失败 0 个" in info_logs[1]
+    assert "其中 1 个订阅正在处理中，本次没有重复检查" in info_logs[1]
+    assert "另有 1 个订阅已停止" in info_logs[1]
+    assert "run_id" not in info_logs[1]
     assert "订阅成功" not in "\n".join(info_logs)
     assert chain._subscription_execution_admission.release(search_lease) is True
 
@@ -293,9 +288,8 @@ def test_match_summary_distinguishes_ttl_timeout(monkeypatch) -> None:
     chain.match({"site.example": [object()]})
 
     assert len(info_logs) == 2
-    assert "task_skipped=1" in info_logs[1]
-    assert "cancelled=0" in info_logs[1]
-    assert "ttl_timeouts=1" in info_logs[1]
+    assert "这次未检查 1 个" in info_logs[1]
+    assert "另有 1 个订阅检查时间过长，已停止" in info_logs[1]
 
 
 def test_match_logs_subscription_admission_release_failure(monkeypatch) -> None:
@@ -323,5 +317,5 @@ def test_match_logs_subscription_admission_release_failure(monkeypatch) -> None:
 
     chain.match({"site.example": [object()]})
 
-    assert any("订阅准入释放失败: operation=match subscription_id=17" in item for item in error_logs)
-    assert "release_failures=1" in info_logs[-1]
+    assert any("订阅 17 的搜索状态没有正常恢复，系统稍后会继续检查" in item for item in error_logs)
+    assert "另有 1 个订阅的搜索状态没有正常恢复" in info_logs[-1]

tests/test_subscription_search_governance.py

@@ -82,6 +82,7 @@ def _chain(tmp_path, subscribes: list[SubscriptionSnapshot]):
     chain._match_lock = _ForbiddenLock()
     chain._search_queue_lock = threading.Lock()
     chain._subscription_execution_admission = SubscriptionExecutionAdmission()
+    chain.messagehelper = Mock()
     return chain
 
 
@@ -179,11 +180,9 @@ def test_successful_sites_remain_available_to_next_due_subscription(tmp_path, mo
 
     assert calls == [(40, 11), (40, 12), (41, 11), (41, 12)]
     assert chain.get_search_batch(batch_id).state == "completed"
-    assert "sites=2" in info_logs[-1]
-    assert "site_requests=4" in info_logs[-1]
-    assert "site_failures=0" in info_logs[-1]
-    assert "site_cooldown_skips=0" in info_logs[-1]
-    assert "candidates=4" in info_logs[-1]
+    assert "访问 2 个站点" in info_logs[-1]
+    assert "发出 4 次请求" in info_logs[-1]
+    assert "找到 4 个资源" in info_logs[-1]
 
 
 def test_fallback_queue_executes_without_match_global_lock(tmp_path, monkeypatch):
@@ -280,12 +279,12 @@ def test_site_budget_conflict_requeues_task_without_batch_failure(tmp_path, monk
         assert task.state == "queued"
         assert task.phase == "waiting_site_budget"
         assert task.available_at == retry_at
-        assert task.last_error is None
+        assert task.last_error == "站点暂时忙，系统会自动继续搜索"
     assert chain.subscription_search_repository.claim_next(owner="worker-after-retry") is None
 
 
-def test_search_logs_one_bounded_start_and_finish_summary(tmp_path, monkeypatch):
-    """Search INFO 只保留轮次摘要，并携带任务终态与耗时字段。"""
+def test_search_logs_one_readable_start_and_finish_summary(tmp_path, monkeypatch):
+    """Search INFO 只保留用户能直接读懂的开始和结束摘要。"""
     subscribes = [_subscribe(50), _subscribe(51)]
     chain = _chain(tmp_path, subscribes)
     _make_tasks_ready(monkeypatch)
@@ -305,19 +304,14 @@ def test_search_logs_one_bounded_start_and_finish_summary(tmp_path, monkeypatch)
 
     assert batch_id
     assert len(info_logs) == 2
-    assert info_logs[0].startswith("订阅治理轮次开始: operation=search ")
-    assert f"batch_id={batch_id}" in info_logs[0]
-    assert "subscriptions=2" in info_logs[0]
-    assert info_logs[1].startswith("订阅治理轮次结束: operation=search ")
-    assert "state=failed" in info_logs[1]
-    assert "processed=2" in info_logs[1]
-    assert "task_completed=1" in info_logs[1]
-    assert "task_failed=1" in info_logs[1]
-    assert "admission_conflicts=0" in info_logs[1]
-    assert "ttl_timeouts=0" in info_logs[1]
-    assert "site_requests=0" in info_logs[1]
-    assert "duration_ms=" in info_logs[1]
-    assert "订阅成功" not in "\n".join(info_logs)
+    assert info_logs[0] == "开始订阅定时检查，共 2 个订阅。"
+    assert info_logs[1].startswith("订阅定时检查结束：部分订阅没有完成。")
+    assert "本次处理 2/2 个" in info_logs[1]
+    assert "完成 1 个" in info_logs[1]
+    assert "失败 1 个" in info_logs[1]
+    assert "发出 0 次请求" in info_logs[1]
+    assert "用时 " in info_logs[1]
+    assert "operation=" not in "\n".join(info_logs)
 
 
 def test_search_logs_subscription_admission_release_failure(tmp_path, monkeypatch):
@@ -343,8 +337,8 @@ def test_search_logs_subscription_admission_release_failure(tmp_path, monkeypatc
     with patch("app.chain.subscribe.search.SearchChain", return_value=Mock()):
         chain.search(state="R")
 
-    assert any("订阅准入释放失败: operation=search subscription_id=52" in item for item in error_logs)
-    assert "release_failures=1" in info_logs[-1]
+    assert any("订阅《治理电影 52》的搜索状态没有正常恢复" in item for item in error_logs)
+    assert "另有 1 个搜索状态没有正常恢复" in info_logs[-1]
 
 
 def test_queued_search_logs_failed_finish_summary_when_callback_raises(tmp_path, monkeypatch):
@@ -361,9 +355,7 @@ def test_queued_search_logs_failed_finish_summary_when_callback_raises(tmp_path,
         )
 
     assert len(info_logs) == 2
-    assert info_logs[-1].startswith("订阅治理轮次结束: operation=search ")
-    assert "state=failed" in info_logs[-1]
-    assert "round_failed=1" in info_logs[-1]
+    assert info_logs[-1].startswith("订阅定时检查结束：部分订阅没有完成。")
 
 
 def test_resume_search_logs_failed_finish_summary_when_drain_raises(tmp_path, monkeypatch):
@@ -380,11 +372,8 @@ def test_resume_search_logs_failed_finish_summary_when_drain_raises(tmp_path, mo
     with pytest.raises(RuntimeError, match="claim failed"):
         chain.resume_search_queue()
 
-    assert len(info_logs) == 2
-    assert info_logs[-1].startswith("订阅治理轮次结束: operation=search ")
-    assert "source=resume" in info_logs[-1]
-    assert "state=failed" in info_logs[-1]
-    assert "round_failed=1" in info_logs[-1]
+    assert len(info_logs) == 1
+    assert info_logs[-1].startswith("等待中的订阅搜索结束：部分订阅没有完成。")
 
 
 def test_swallowed_indexer_failure_marks_task_and_batch_failed_but_continues(tmp_path, monkeypatch):
@@ -457,10 +446,9 @@ def test_swallowed_indexer_failure_marks_task_and_batch_failed_but_continues(tmp
     assert batch.failed_count == 1
     assert "Flaky" in batch.last_error
     assert "HTTP 429" in batch.last_error
-    assert "sites=2" in info_logs[-1]
-    assert "site_requests=2" in info_logs[-1]
-    assert "site_failures=1" in info_logs[-1]
-    assert "cooldown_seconds=900.0" in info_logs[-1]
+    assert "失败 1 个" in info_logs[-1]
+    assert "访问 2 个站点" in info_logs[-1]
+    assert "发出 2 次请求" in info_logs[-1]
 
 
 def test_same_subscription_conflict_is_skipped_without_waiting(tmp_path, monkeypatch):
@@ -487,12 +475,116 @@ def test_same_subscription_conflict_is_skipped_without_waiting(tmp_path, monkeyp
     assert batch.state == "skipped"
     assert batch.finished_count == 0
     assert batch.skipped_count == 1
-    assert batch.last_error == "同一订阅正在由其他通道处理，本轮搜索已跳过"
-    assert "task_skipped=1" in info_logs[-1]
-    assert "admission_conflicts=1" in info_logs[-1]
+    assert batch.last_error == "这个订阅正在处理，本次自动检查无需重复执行"
+    assert "这次未搜索 1 个" in info_logs[-1]
     assert chain._subscription_execution_admission.release(match_lease) is True
 
 
+def test_manual_search_waits_for_same_subscription_then_continues(tmp_path, monkeypatch):
+    """手动搜索遇到同一订阅正在处理时应稍后自动继续，而不是结束为跳过。"""
+    subscribe = _subscribe(61)
+    chain = _chain(tmp_path, [subscribe])
+    _make_tasks_ready(monkeypatch)
+    process = Mock(return_value=subscribe)
+    monkeypatch.setattr(chain, "_process_search_subscription", process)
+    match_lease = chain._subscription_execution_admission.try_acquire(
+        subscription_id=subscribe.id,
+        operation="match",
+        ttl_seconds=60,
+    )
+    assert match_lease is not None
+
+    with patch("app.chain.subscribe.search.SearchChain", return_value=Mock()):
+        batch_id = chain.search(sid=subscribe.id, state=None, manual=True)
+
+    batch = chain.get_search_batch(batch_id)
+    assert batch.state == "queued"
+    process.assert_not_called()
+    with chain.subscription_search_repository._session_factory() as session:
+        task = session.query(SubscriptionSearchTask).filter_by(
+            subscription_id=subscribe.id,
+        ).one()
+        assert task.phase == "waiting_subscription"
+        assert task.last_error == "这个订阅正在处理，结束后会自动继续搜索"
+        task.available_at = "1970-01-01T00:00:00+00:00"
+        session.commit()
+
+    assert chain._subscription_execution_admission.release(match_lease) is True
+    with patch("app.chain.subscribe.search.SearchChain", return_value=Mock()):
+        chain.resume_search_queue(limit=1)
+
+    process.assert_called_once()
+    assert chain.get_search_batch(batch_id).state == "completed"
+
+
+def test_recent_new_subscription_is_scheduled_instead_of_skipped(tmp_path, monkeypatch):
+    """新订阅的一分钟编辑窗口应保留任务并在到时后自动继续。"""
+    subscribe = replace(
+        _subscribe(62),
+        state="N",
+        date=datetime.now().strftime("%Y-%m-%d %H:%M:%S"),
+    )
+    chain = _chain(tmp_path, [subscribe])
+    _make_tasks_ready(monkeypatch)
+    process = Mock(side_effect=AssertionError("编辑等待期内不应开始搜索"))
+    monkeypatch.setattr(chain, "_process_search_subscription", process)
+
+    with patch("app.chain.subscribe.search.SearchChain", return_value=Mock()):
+        batch_id = chain.search(state="N")
+
+    batch = chain.get_search_batch(batch_id)
+    assert batch.state == "queued"
+    assert batch.skipped_count == 0
+    with chain.subscription_search_repository._session_factory() as session:
+        task = session.query(SubscriptionSearchTask).filter_by(
+            subscription_id=subscribe.id,
+        ).one()
+        assert task.phase == "scheduled"
+        assert task.last_error == "订阅刚刚创建，保存好设置后会自动开始搜索"
+        assert task.available_at > datetime.now(timezone.utc).isoformat(timespec="seconds")
+    process.assert_not_called()
+
+
+def test_new_subscription_post_commit_creates_scheduled_search(tmp_path):
+    """订阅保存后应立即写入自动搜索计划，无需等待五分钟扫描。"""
+    subscribe = replace(
+        _subscribe(64),
+        state="N",
+        date=datetime.now().strftime("%Y-%m-%d %H:%M:%S"),
+    )
+    chain = _chain(tmp_path, [subscribe])
+
+    batch_id = chain._SubscribeChain__queue_new_subscription_search(subscribe.id)
+
+    assert batch_id
+    assert chain.get_search_batch(batch_id).state == "queued"
+    with chain.subscription_search_repository._session_factory() as session:
+        task = session.query(SubscriptionSearchTask).filter_by(
+            subscription_id=subscribe.id,
+        ).one()
+        assert task.source == "new"
+        assert task.phase == "scheduled"
+        assert task.available_at > datetime.now(timezone.utc).isoformat(timespec="seconds")
+
+
+def test_manual_search_bypasses_new_subscription_edit_wait(tmp_path, monkeypatch):
+    """用户主动点击搜索时不受新订阅编辑等待时间限制。"""
+    subscribe = replace(
+        _subscribe(63),
+        date=datetime.now().strftime("%Y-%m-%d %H:%M:%S"),
+    )
+    chain = _chain(tmp_path, [subscribe])
+    _make_tasks_ready(monkeypatch)
+    process = Mock(return_value=subscribe)
+    monkeypatch.setattr(chain, "_process_search_subscription", process)
+
+    with patch("app.chain.subscribe.search.SearchChain", return_value=Mock()):
+        batch_id = chain.search(sid=subscribe.id, state=None, manual=True)
+
+    process.assert_called_once()
+    assert chain.get_search_batch(batch_id).state == "completed"
+
+
 def test_paused_subscription_is_skipped_after_admission_refresh(tmp_path, monkeypatch):
     """准入后重新读取到暂停订阅时不得开始搜索或伪造完成。"""
     subscribe = _subscribe(9)
@@ -511,7 +603,7 @@ def test_paused_subscription_is_skipped_after_admission_refresh(tmp_path, monkey
     assert batch.state == "skipped"
     assert batch.finished_count == 0
     assert batch.skipped_count == 1
-    assert batch.last_error == "订阅已暂停，本轮搜索已跳过"
+    assert batch.last_error == "订阅已暂停，这次没有搜索"
 
 
 def test_cleanup_failures_cannot_leak_subscription_admission(tmp_path, monkeypatch):
@@ -699,9 +791,8 @@ def test_ttl_expiry_after_normal_return_marks_task_and_batch_failed(tmp_path, mo
     assert batch.finished_count == 0
     assert batch.failed_count == 1
     assert batch.cancelled_count == 0
-    assert batch.last_error == "订阅执行已超过协作截止时间"
-    assert "task_failed=1" in info_logs[-1]
-    assert "ttl_timeouts=1" in info_logs[-1]
+    assert batch.last_error == "这次搜索用时过长，已停止，可稍后重试"
+    assert "失败 1 个" in info_logs[-1]
 
 
 def test_ttl_expiry_after_download_started_completes_with_actual_result(tmp_path, monkeypatch):
@@ -758,4 +849,4 @@ def test_site_budget_ttl_expiry_marks_task_and_batch_failed(tmp_path, monkeypatc
     assert batch.finished_count == 0
     assert batch.failed_count == 1
     assert batch.cancelled_count == 0
-    assert batch.last_error == "订阅执行已超过协作截止时间"
+    assert batch.last_error == "这次搜索用时过长，已停止，可稍后重试"

tests/test_subscription_search_queue.py

@@ -40,7 +40,9 @@ def test_search_queue_coalesces_active_subscription_and_raises_priority(tmp_path
     assert manual.created_count == 0
     assert manual.coalesced_count == 1
     assert manual.batch.state == "completed"
+    assert manual.active_batch_ids == (scheduled.batch.batch_id,)
     assert first.subscription_id == 1
+    assert first.source == "manual"
     assert first.priority == 100
     assert second.subscription_id == 2
     assert first.task_id != second.task_id
@@ -82,6 +84,39 @@ def test_search_queue_claims_each_subscription_only_after_its_available_at(tmp_p
     assert accelerated.priority == 100
 
 
+def test_manual_search_promotes_scheduled_new_subscription(tmp_path):
+    """用户主动搜索应立即唤醒仍在编辑等待期的新订阅任务。"""
+    repository, engine = _repository(tmp_path)
+    later_at = (datetime.now(timezone.utc) + timedelta(minutes=1)).isoformat(timespec="seconds")
+    automatic = repository.enqueue(
+        subscription_ids=(22,),
+        source="new",
+        priority=50,
+        available_at_by_subscription={22: later_at},
+    )
+
+    with Session(engine) as session:
+        scheduled = session.execute(
+            select(SubscriptionSearchTask).where(SubscriptionSearchTask.subscription_id == 22)
+        ).scalar_one()
+        assert scheduled.phase == "scheduled"
+
+    manual = repository.enqueue(
+        subscription_ids=(22,),
+        source="manual",
+        priority=120,
+        available_at_by_subscription={22: "1970-01-01T00:00:00+00:00"},
+    )
+    claimed = repository.claim_next(owner="worker-manual")
+
+    assert manual.created_count == 0
+    assert manual.active_batch_ids == (automatic.batch.batch_id,)
+    assert claimed is not None
+    assert claimed.source == "manual"
+    assert claimed.priority == 120
+    assert claimed.phase == "matching"
+
+
 def test_search_queue_recovers_expired_lease_with_same_task_identity(tmp_path):
     """进程遗留的过期 running 任务应以新 token 恢复且 attempt 单调递增。"""
     repository, engine = _repository(tmp_path)
@@ -269,8 +304,8 @@ def test_search_queue_aggregates_skipped_tasks_without_marking_success(tmp_path)
     assert batch.last_error == "同一订阅正在由其他通道处理，本轮搜索已跳过"
 
 
-def test_search_queue_ages_old_fallback_ahead_of_new_manual_work(tmp_path):
-    """手工任务可优先，但等待超过公平窗口的兜底任务不得持续饥饿。"""
+def test_search_queue_keeps_manual_work_ahead_of_aged_fallback(tmp_path):
+    """用户主动搜索始终先于定时检查，避免点击后长时间没有反馈。"""
     repository, engine = _repository(tmp_path)
     repository.enqueue(subscription_ids=(8,), source="fallback", priority=10)
     aged_at = (datetime.now(timezone.utc) - timedelta(minutes=16)).isoformat(timespec="seconds")
@@ -285,5 +320,5 @@ def test_search_queue_ages_old_fallback_ahead_of_new_manual_work(tmp_path):
 
     claimed = repository.claim_next(owner="worker-a")
 
-    assert claimed.subscription_id == 8
-    assert claimed.source == "fallback"
+    assert claimed.subscription_id == 9
+    assert claimed.source == "manual"

tests/test_subscription_site_budget.py

@@ -35,6 +35,7 @@ def _repository(tmp_path):
 def test_site_budget_allows_one_inflight_per_site_and_independent_sites(tmp_path):
     """同站点第二个调用必须等待，不同站点可立即并行。"""
     repository, _engine = _repository(tmp_path)
+    before_retry = datetime.now(timezone.utc)
 
     first = repository.claim_site(site_id=1, owner="worker-a", lease_seconds=900)
     same_site = repository.claim_site(site_id=1, owner="worker-b", lease_seconds=900)
@@ -43,6 +44,8 @@ def test_site_budget_allows_one_inflight_per_site_and_independent_sites(tmp_path
     assert first.acquired is True
     assert same_site.acquired is False
     assert same_site.lease_token is None
+    assert same_site.wait_reason == "busy"
+    assert datetime.fromisoformat(same_site.retry_at) <= before_retry + timedelta(seconds=11)
     assert other_site.acquired is True