diff --git a/backend/app/ledger/view.py b/backend/app/ledger/view.py
index 81492c2b..5ce136a6 100644
--- a/backend/app/ledger/view.py
+++ b/backend/app/ledger/view.py
@@ -10,6 +10,7 @@ issued 是这次运行按口径出具过的正式运行，reconciled 是对过
 """
 from __future__ import annotations
 
+from datetime import datetime, timezone
 from typing import Any
 
 from sqlalchemy import select
@@ -23,6 +24,10 @@ from app.ledger import store
 #: state：matched 比对过；pending 已排队按需比对；ineligible 这次运行不比对（reason 写为什么）；unreadable 比对记录读不出
 STATES = ("matched", "pending", "ineligible", "unreadable")
 
+#: 成功结束、还没封存的运行，结束后这么多秒内答 pending（runner 落库之后马上封存，正常是几秒）。过了还没封存的是封存出了错、
+#: 或者是封存之前的老数据：再答 pending，页面就一直转圈、等满两分钟（store/ledger.ts 的等待上限）再取还是 pending——照实说没封存
+SEAL_GRACE_S = 120
+
 
 class _Hide:
     """按这次运行查到过的遮罩值遮：hide(值) 这个值给不给，shown(文字) 文字里的遮罩值换成「已遮罩」。"""
@@ -92,8 +97,6 @@ def _labels_view(content: dict[str, Any], hider: _Hide, names: dict[str, str]) -
 
 def _stamp(recorded: store.Recorded) -> dict[str, Any]:
     """比对的来历：哪份工件、哪一版算法、谁触发的、什么时候追加的；事后补比的写明按现在的口径回看。"""
-    from datetime import datetime, timezone
-
     at = datetime.fromtimestamp(recorded.ts, timezone.utc).isoformat(timespec="seconds") \
         if isinstance(recorded.ts, (int, float)) else None
     return {"artifact": recorded.artifact, "algo": recorded.algo, "trigger": recorded.trigger, "retro": recorded.retro,
@@ -102,6 +105,14 @@ def _stamp(recorded: store.Recorded) -> dict[str, Any]:
                 else ledger_labels.RETRO_NOTE} if recorded.retro else {})}
 
 
+def _just_finished(finished_at: datetime | None) -> bool:
+    """运行是不是刚结束（SEAL_GRACE_S 以内）。没记结束时间的不算；库里读回 naive 的按 UTC 算（同 db/base.UTCDateTime）。"""
+    if finished_at is None:
+        return False
+    at = finished_at if finished_at.tzinfo is not None else finished_at.replace(tzinfo=timezone.utc)
+    return (datetime.now(timezone.utc) - at).total_seconds() <= SEAL_GRACE_S
+
+
 async def run_view(run_id: str) -> dict[str, Any]:
     """GET /api/runs/{run_id}/ledger 的答复：标签、整题判断、每条查询的比对、能做的动作。没比对过、能比的运行这时排上按需比对
     （trigger=on_demand），答复 state=pending；不比的照实说为什么。"""
@@ -117,9 +128,10 @@ async def run_view(run_id: str) -> dict[str, Any]:
     if recorded is None:
         why = ledger_match.ineligible(run)
         if why == ledger_match.NOT_SEALED and run.status == "succeeded" \
-                and ledger_match.wanted(run.run_class, run.origin_kind):
+                and ledger_match.wanted(run.run_class, run.origin_kind) and _just_finished(run.finished_at):
             # runner 先发 run.finished、落库，再封存、再排上比对（engine/runner._finalize）：页面收到 run.finished 立刻来取，
-            # 这一下还没封存。这不是「不比对」，是封存完就比——照实答 pending，页面按退避接着取；这里不另排按需比对
+            # 这一下还没封存。这不是「不比对」，是封存完就比——照实答 pending，页面按退避接着取；这里不另排按需比对。
+            # 只限刚结束的那一会儿（SEAL_GRACE_S），早就结束、一直没封存的照实答 ineligible
             return {**base, "state": "pending"}
         sealed = await _sealed(run_id) if why is None else None
         if sealed is not None and not sealed.queries:
diff --git a/backend/tests/test_ledger_api.py b/backend/tests/test_ledger_api.py
index e70a6561..fceeb799 100644
--- a/backend/tests/test_ledger_api.py
+++ b/backend/tests/test_ledger_api.py
@@ -102,11 +102,14 @@ async def test_a_run_that_just_finished_and_is_about_to_be_sealed_is_pending(cli
     await ledger_match.drain()
     from app.db.models import Run
 
+    from datetime import datetime, timedelta, timezone
+
+    now = datetime.now(timezone.utc)
     async with SessionLocal() as session:
         fresh = Run(workflow_name="刚跑完", status="succeeded", run_class="exploratory", origin_kind="chat", graph={},
-                    input={}, output={})
+                    input={}, output={}, finished_at=now)
         failed = Run(workflow_name="失败了", status="failed", run_class="exploratory", origin_kind="chat", graph={},
-                     input={}, output={})
+                     input={}, output={}, finished_at=now)
         session.add_all([fresh, failed])
         await session.commit()
         fresh_id, failed_id = fresh.id, failed.id
@@ -117,6 +120,32 @@ async def test_a_run_that_just_finished_and_is_about_to_be_sealed_is_pending(cli
     assert (await ledger_of(client, failed_id))["state"] == "ineligible"
 
 
+async def test_a_run_that_finished_long_ago_and_was_never_sealed_says_so(client, monkeypatch):
+    """「刚跑完、马上封存」只是结束后的那一会儿：成功结束了好一阵、一直没封存的（封存时出了错、封存这件事之前的老数据），
+    再答 pending 就是让页面一直转圈、等两分钟、点「再取一次」还是 pending。照实说没封存（ineligible，NOT_SEALED）；
+    结束时间没记的同样不当成刚结束。"""
+    from datetime import datetime, timedelta, timezone
+
+    from app.db.models import Run
+    from app.ledger import view as ledger_view
+
+    old = datetime.now(timezone.utc) - timedelta(seconds=ledger_view.SEAL_GRACE_S + 60)
+    async with SessionLocal() as session:
+        stale = Run(workflow_name="很早以前跑完", status="succeeded", run_class="exploratory", origin_kind="chat",
+                    graph={}, input={}, output={}, finished_at=old)
+        undated = Run(workflow_name="没记结束时间", status="succeeded", run_class="exploratory", origin_kind="chat",
+                      graph={}, input={}, output={})
+        session.add_all([stale, undated])
+        await session.commit()
+        stale_id, undated_id = stale.id, undated.id
+    scheduled: list[str] = []
+    monkeypatch.setattr(ledger_match, "schedule", lambda run_id, **kw: scheduled.append(run_id) or True)
+    for run_id in (stale_id, undated_id):
+        body = await ledger_of(client, run_id)
+        assert body["state"] == "ineligible" and body["reason"] == ledger_match.NOT_SEALED, body
+    assert scheduled == []
+
+
 async def test_runs_that_are_not_compared_say_why(client, monkeypatch):
     row = await S.explore(monkeypatch, S.EQUIVALENT, origin="api", text="本周销售额 [[v:Q1.r0.销售额]]。")
     body = await ledger_of(client, row.id)