Surface failed-turn reason: host.py recovers it from thread_history, dispatch prints it

Real bad-model turn: relay returns 404, codex retries sampling
5 times, turn fails, but app-server's turn/completed event
carries status=failed with NO error and NO text, so worker
never saw any fallbackable signal. The 404 actually lives in
codex's thread_history_1.sqlite (thread_turns.error_json).
host.py Server.run_turn now best-effort recovers that message
(read-only query by thread_id+turn_id, any failure -> None,
never breaks the turn itself), and dispatch CLI prints the
error field (run CLI already printed the full result). Worker
_attempt_error prefers the error field when text is empty, so
a double failure keeps the real reasons, not just status codes.
Tests: host TurnErrorFromHistoryTest (recovery + missing/corrupt
DB -> None), full host + cloud suites pass.

tests/test_host.py

@@ -248,6 +248,53 @@ class RetryAndBudgetTest(unittest.TestCase):
             self.assertIn("t1 failed", Path(notify_log).read_text())
 
 
+class TurnErrorFromHistoryTest(unittest.TestCase):
+    def _server_with_history(self, tmp, error_json):
+        import sqlite3
+        db = os.path.join(tmp, "thread_history_1.sqlite")
+        conn = sqlite3.connect(db)
+        conn.execute(
+            "CREATE TABLE thread_turns(thread_id TEXT, turn_id TEXT,"
+            " error_json TEXT)")
+        if error_json is not None:
+            conn.execute(
+                "INSERT INTO thread_turns VALUES('th1', 'tu1', ?)",
+                (error_json,))
+        conn.commit()
+        conn.close()
+        srv = object.__new__(host.Server)
+        srv.codex_home = tmp
+        return srv
+
+    def test_failed_turn_recovers_404_from_history(self):
+        # 事件里 status=failed 但 error 为空时，从 thread_history
+        # 库里按 thread_id + turn_id 捞回 404 原因（主机实测回归：
+        # 坏模型回合的 404 只在库里，事件里没有，导致 worker
+        # 永远看不到可 fallback 信号）。
+        import json as _json
+        with tempfile.TemporaryDirectory() as tmp:
+            err = _json.dumps({
+                "message": "unexpected status 404 Not Found: Model "
+                           "\"gpt-9.9-nonexistent\" is not available",
+                "codexErrorInfo": {"httpConnectionFailed":
+                                   {"httpStatusCode": 404}}})
+            srv = self._server_with_history(tmp, err)
+            got = srv._turn_error_from_history("th1", "tu1")
+            self.assertIn("404", got)
+            self.assertIn("gpt-9.9-nonexistent", got)
+
+    def test_missing_history_returns_none_without_crashing(self):
+        # 库不存在、行不存在、JSON 损坏都只返回 None，不抛错。
+        with tempfile.TemporaryDirectory() as tmp:
+            srv = self._server_with_history(tmp, None)
+            self.assertIsNone(srv._turn_error_from_history("th1", "tu1"))
+            self.assertIsNone(srv._turn_error_from_history("th1", None))
+            self.assertIsNone(srv._turn_error_from_history(None, "tu1"))
+        with tempfile.TemporaryDirectory() as tmp:
+            srv = self._server_with_history(tmp, "not-json{")
+            self.assertIsNone(srv._turn_error_from_history("th1", "tu1"))
+
+
 class StateVersionTest(unittest.TestCase):
     def test_version_increments_and_conflict_logged(self):
         with tempfile.TemporaryDirectory() as tmp: