长任务 P0 三件：断点进库、错误分类+熔断、Run/Step 时间线

断点进库：task_checkpoints 表（PG 新库 + 老库幂等补表 + 迁移清单
对齐）；worker 在 goal 任务执行期间每 30 秒把磁盘 host-state 断点
镜像进库，终局再补一次；断点续跑认领前，若 host-state 条目丢失
但库里有断点，先按库断点重建最小条目再跑，库里也没有断点时不
造假、走原有拒绝路径。

错误分类+熔断：sandbox 四处解析加永久错误分支（KeyError、配置
缺失等无结论异常退出判 failed 死信，不再判 interrupted；host 已
明言 interrupted 的不受影响，未知错误不分类）；worker 自动恢复
加退避（10 秒起指数退避、封顶 300 秒）与无进展熔断（库断点轮次
连续 3 次不变即标 auto_resume_stalled 等人工）。

时间线：task_steps 表 + host 每轮写 host-steps 步骤文件 + worker
导入；GET /tasks/{id}/timeline 一次返回 steps 与 checkpoints；
认领、暂停、自动恢复、终局都记步骤。旧任务返回空数组，不编造。

单测：cloud/tests/test_p0_three.py 14 项（分类、熔断、镜像重建、
步骤导入去重、timeline 接口）。

cloud/server.py

@@ -1126,6 +1126,19 @@ def migrate_tasks_table(conn):
         "interrupted_reason": "TEXT",
         "auto_resume_count": "INTEGER DEFAULT 0",
         "no_time_limit": "INTEGER DEFAULT 0",
+        # P0 三件：自动恢复退避与无进展熔断（见 worker
+        # auto_resume_interrupted）：resume_after 是下次允许
+        # 重排队的时间戳（指数退避）；auto_resume_stall_count
+        # 记连续无进展重排队次数（到 3 即停、标
+        # auto_resume_stalled 等人工）；checkpoint_round_at_last_
+        # resume 记上次重排队时库中断点轮次（无进展判定的
+        # 基线，NULL = 从未记过）；last_error_signature 记
+        # 最近一次终局的错误签名（permanent/transient/None，
+        # 供排障，不作恢复判定——判定在 sandbox 终态里）。
+        "resume_after": "DOUBLE PRECISION DEFAULT 0",
+        "auto_resume_stall_count": "INTEGER DEFAULT 0",
+        "checkpoint_round_at_last_resume": "INTEGER",
+        "last_error_signature": "TEXT",
     }
     for name, coltype in new_cols.items():
         if name not in cols:
@@ -1204,6 +1217,40 @@ def _attachments_map(conn, task_ids):
     return out
 
 
+def _ensure_task_checkpoints_table(conn):
+    """幂等建断点表（P0 三件其一）：server init_db（PG 与 SQLite
+    两支）与 worker 迁移兜底共用。DOUBLE PRECISION 在两后端都可用。
+    """
+    conn.execute(
+        "CREATE TABLE IF NOT EXISTS task_checkpoints("
+        "task_id TEXT, "
+        "round_no INTEGER, "
+        "thread_id TEXT, "
+        "goal_status TEXT, "
+        "state_sha TEXT, "
+        "verify_out TEXT, "
+        "created_at DOUBLE PRECISION, "
+        "PRIMARY KEY(task_id, round_no))")
+
+
+def _ensure_task_steps_table(conn):
+    """幂等建步骤表（P0 三件其三）：timeline 接口直读此表。
+    id 由写入方给 uuid hex（两后端都不依赖自增语义）。
+    """
+    conn.execute(
+        "CREATE TABLE IF NOT EXISTS task_steps("
+        "id TEXT PRIMARY KEY, "
+        "task_id TEXT, "
+        "step_no INTEGER, "
+        "kind TEXT, "
+        "summary TEXT, "
+        "state_sha TEXT, "
+        "created_at DOUBLE PRECISION)")
+    conn.execute(
+        "CREATE INDEX IF NOT EXISTS idx_task_steps_task"
+        " ON task_steps(task_id, step_no)")
+
+
 def _ensure_followups_table(conn):
     """幂等建后续消息表：委托给 peer_messages（两边同一份 DDL，
     server init_db、各 followup 接口与 worker 投递路径共用）。
@@ -1530,6 +1577,8 @@ def init_db(db_path):
             _backfill_default_quotas(conn)
             _ensure_gateway_keys(conn)
             _ensure_uploads_tables(conn)
+            _ensure_task_checkpoints_table(conn)
+            _ensure_task_steps_table(conn)
             conn.commit()
             return
         conn.execute(
@@ -1564,6 +1613,8 @@ def init_db(db_path):
         _ensure_quotas_table(conn)
         _backfill_default_quotas(conn)
         _ensure_uploads_tables(conn)
+        _ensure_task_checkpoints_table(conn)
+        _ensure_task_steps_table(conn)
         # worker 每轮 poll 都按 status 取任务（见 load_tasks 的
         # WHERE status IN (...)），无索引时是全表扫描；任务表随
         # 时间变大后索引是必须的。IF NOT EXISTS 保证老库幂等。
@@ -2099,6 +2150,8 @@ class CloudControlPlaneHandler(BaseHTTPRequestHandler):
                 self._create_followup(user_id, rest[: -len("/followups")])
             elif rest.endswith("/followups") and method == "GET":
                 self._list_followups(user_id, rest[: -len("/followups")])
+            elif rest.endswith("/timeline") and method == "GET":
+                self._task_timeline(user_id, rest[: -len("/timeline")])
             elif rest.endswith("/chat-history") and method == "GET":
                 self._task_chat_history(
                     user_id, rest[: -len("/chat-history")])
@@ -4602,6 +4655,49 @@ class CloudControlPlaneHandler(BaseHTTPRequestHandler):
         self._send_json(200, {
             "followups": [_followup_to_dict(r) for r in rows]})
 
+    def _task_timeline(self, user_id, task_id):
+        """GET /tasks/{id}/timeline（P0 三件其三）：一个任务一张
+        时间线。steps 直读 task_steps（claim/resume/round/final），
+        checkpoints 直读 task_checkpoints（每轮断点快照）；旧
+        任务（部署前跑完、无步骤记录）返回空数组，不编造。
+        """
+        row = self._fetch_task(task_id)
+        if row is None:
+            self._send_json(404, {"error": "task not found"})
+            return
+        task = _task_to_dict(row)
+        with _auth_store(self.server.db_path) as store:
+            if not store.check_permission(
+                    user_id, task["project_id"], "read"):
+                self._send_json(403, {"error": "forbidden"})
+                return
+        conn = self._db()
+        try:
+            _ensure_task_checkpoints_table(conn)
+            _ensure_task_steps_table(conn)
+            step_rows = conn.execute(
+                "SELECT step_no, kind, summary, state_sha, created_at"
+                " FROM task_steps WHERE task_id = ?"
+                " ORDER BY step_no, created_at, id",
+                (task_id,)).fetchall()
+            cp_rows = conn.execute(
+                "SELECT round_no, thread_id, goal_status, state_sha,"
+                " verify_out, created_at FROM task_checkpoints"
+                " WHERE task_id = ? ORDER BY round_no",
+                (task_id,)).fetchall()
+        finally:
+            conn.close()
+        self._send_json(200, {
+            "task_id": task_id,
+            "steps": [{"step_no": r[0], "kind": r[1], "summary": r[2],
+                       "state_sha": r[3], "created_at": r[4]}
+                      for r in step_rows],
+            "checkpoints": [{"round_no": r[0], "thread_id": r[1],
+                             "goal_status": r[2], "state_sha": r[3],
+                             "verify_out": r[4], "created_at": r[5]}
+                            for r in cp_rows],
+        })
+
     def _task_chat_history(self, user_id, task_id):
         """GET /tasks/{id}/chat-history：真正的交替历史，一次性
         按时间交错拼好返回，前端不再猜顺序。
@@ -6541,6 +6637,21 @@ class CloudControlPlaneHandler(BaseHTTPRequestHandler):
                 " updated_at = ? WHERE id = ?",
                 (now, task_id),
             )
+            # P0 三件其三：时间线 pause 步骤（best-effort）。
+            try:
+                _ensure_task_steps_table(conn)
+                _max = conn.execute(
+                    "SELECT MAX(step_no) FROM task_steps"
+                    " WHERE task_id = ?", (task_id,)).fetchone()
+                conn.execute(
+                    "INSERT INTO task_steps(id, task_id, step_no, kind,"
+                    " summary, state_sha, created_at)"
+                    " VALUES(?, ?, ?, 'pause', '用户暂停', NULL, ?)",
+                    (uuid.uuid4().hex, task_id,
+                     ((_max[0] or 0) + 1 if _max and _max[0] else 1),
+                     now))
+            except Exception:
+                pass
             conn.commit()
             task = _task_to_dict(
                 conn.execute(