diff --git a/app/application/site/throttle.py b/app/application/site/throttle.py
index b1171e902..e3d93f614 100644
--- a/app/application/site/throttle.py
+++ b/app/application/site/throttle.py
@@ -6,6 +6,7 @@ import asyncio
 import math
 import time
 from collections.abc import Mapping
+from datetime import datetime, timedelta
 from typing import Optional
 
 from app.application.site.protocol import site_domain
@@ -15,6 +16,9 @@ from app.runtime.rate import RequestRateLimiter
 from app.schemas.exception import RateLimitExceededException
 
 _site_requests = RequestRateLimiter()
+# 本地额度最多原地等待的秒数；更长的间隔或计数窗口交还调用方延后，
+# 避免订阅搜索消费者等串行执行者在单个请求上休眠到下一个额度。
+_MAX_LOCAL_WAIT_SECONDS = 60.0
 
 
 def site_request_config(url: str, site_id: Optional[int] = None) -> dict[str, object]:
@@ -56,7 +60,7 @@ def _positive_number(value: object) -> float:
 
 def _request_delay(site: Mapping[str, object]) -> float:
     """
-    原子申请一次站点请求，远端冷却期间直接返回原始原因。
+    原子申请一次站点请求，远端冷却或本地等待超过上限时直接抛出原因，不消耗额度。
     """
     interval = _positive_number(site.get("limit_seconds"))
     delay, reason = _site_requests.acquire(
@@ -65,15 +69,29 @@ def _request_delay(site: Mapping[str, object]) -> float:
     )
     if reason:
         raise RateLimitExceededException(reason)
+    if delay > _MAX_LOCAL_WAIT_SECONDS:
+        raise RateLimitExceededException(_local_limit_reason(delay))
     return delay
 
 
+def _local_limit_reason(delay: float) -> str:
+    """
+    描述超过原地等待上限的本地额度，保留限流语义和剩余秒数供下载冷却识别。
+
+    :param delay: 距离下一次本地许可的秒数
+    :return: 给用户看的限流说明
+    """
+    seconds = math.ceil(delay)
+    ready_at = (datetime.now() + timedelta(seconds=seconds)).strftime("%H:%M")
+    return f"站点限流：站点设置的访问间隔或次数未到，还需等待 {seconds} 秒（约 {ready_at} 后可再访问）"
+
+
 def wait_for_site_request(site: Mapping[str, object]) -> None:
     """
     在同步请求前等待配置的间隔和次数窗口，唤醒后重新检查许可。
 
     :param site: 本次请求所属站点的流控配置
-    :raises RateLimitExceededException: 站点仍处于远端明确要求的冷却期
+    :raises RateLimitExceededException: 站点仍处于远端明确要求的冷却期，或本地额度需等待超过上限
     """
     while delay := _request_delay(site):
         time.sleep(delay)
@@ -84,7 +102,7 @@ async def async_wait_for_site_request(site: Mapping[str, object]) -> None:
     异步等待同一站点额度，取消等待不会提前消耗请求次数。
 
     :param site: 本次请求所属站点的流控配置
-    :raises RateLimitExceededException: 站点仍处于远端明确要求的冷却期
+    :raises RateLimitExceededException: 站点仍处于远端明确要求的冷却期，或本地额度需等待超过上限
     """
     while delay := _request_delay(site):
         await asyncio.sleep(delay)
diff --git a/tests/test_site_request_throttle.py b/tests/test_site_request_throttle.py
index a8eb20830..16f5cda97 100644
--- a/tests/test_site_request_throttle.py
+++ b/tests/test_site_request_throttle.py
@@ -19,6 +19,7 @@ from app.adapters.cache.backends import FileBackend
 from app.application.site.observation import capture_site_search_observation
 from app.chain.download import DownloadChain
 from app.chain.search import provider
+from app.domain import torrent as torrent_rules
 from app.domain.context import TorrentInfo
 from app.modules.indexer import IndexerModule
 from app.runtime.rate import RequestRateLimiter
@@ -321,3 +322,67 @@ def test_remote_cooldown_cannot_be_shortened(boundary):
     assert throttle.site_request_cooldown(site) == 590
     with pytest.raises(RateLimitExceededException, match="600 seconds"):
         throttle.wait_for_site_request(site)
+
+
+def test_long_local_interval_defers_instead_of_sleeping(boundary):
+    """
+    超过原地等待上限的访问间隔立即返回限流原因，不休眠、不占用额度，到期后恢复。
+    """
+    site = {"id": 42, "limit_seconds": 3600}
+    throttle.wait_for_site_request(site)
+    boundary.clock.now += 10
+
+    with pytest.raises(RateLimitExceededException, match="还需等待 3590 秒") as error:
+        throttle.wait_for_site_request(site)
+
+    assert torrent_rules.rate_limit_cooldown(str(error.value)) == 3600
+    assert boundary.clock.waits == []
+    boundary.clock.now += 3590
+    throttle.wait_for_site_request(site)
+    assert boundary.clock.waits == []
+
+
+def test_long_count_window_defers_and_short_wait_still_sleeps(boundary):
+    """
+    计数窗口剩余时间超过上限时延后，未超过上限的等待仍按原语义原地完成。
+    """
+    site = {"id": 42, "limit_interval": 3600, "limit_count": 2, "limit_seconds": 60}
+    throttle.wait_for_site_request(site)
+    throttle.wait_for_site_request(site)
+
+    assert boundary.clock.waits == [60]
+    with pytest.raises(RateLimitExceededException, match="还需等待 3540 秒"):
+        throttle.wait_for_site_request(site)
+    assert boundary.clock.waits == [60]
+
+
+def test_torrent_download_after_search_returns_without_waiting(boundary):
+    """
+    搜索刚用掉一小时一次的额度后，同站取种子立即失败并按限流冷却，不阻塞调用线程。
+    """
+    boundary.query.get_sync.side_effect = lambda site_id: Site(id=site_id, limit_seconds=3600)
+    provider._search_site_page(
+        boundary.owner, site={"id": 42, "limit_seconds": 3600}, keyword="Series", media_type=None, page=0,
+    )
+    errors = []
+
+    content = boundary.chain.download_torrent(TorrentInfo(site=42, enclosure=URL, title="Series"),
+                                              on_failure=errors.append)
+
+    assert content == (None, "", [])
+    assert errors and errors[0].startswith("站点限流：站点设置的访问间隔或次数未到")
+    assert DownloadChain._download_failure_ttl(errors[0]) == 3600
+    boundary.port.request.assert_not_called()
+    assert boundary.clock.waits == []
+
+
+@pytest.mark.asyncio
+async def test_async_long_local_interval_defers(boundary, monkeypatch):
+    """
+    异步搜索同样不为超过上限的本地额度长时间挂起。
+    """
+    monkeypatch.setattr(throttle.asyncio, "sleep", AsyncMock(side_effect=AssertionError("不应等待")))
+    site = {"id": 42, "limit_seconds": 3600}
+    await throttle.async_wait_for_site_request(site)
+    with pytest.raises(RateLimitExceededException, match="还需等待 3600 秒"):
+        await throttle.async_wait_for_site_request(site)