fix(exec): retain stale sessions after cleanup failure

Only remove idle exec sessions after process termination succeeds so later cleanup and shutdown paths can retry failed kills.

nanobot/agent/tools/exec_session.py

@@ -370,8 +370,9 @@ class ExecSessionManager:
             if now - session.last_access > self.idle_timeout
         ]
         for session_id in stale:
-            session = self._sessions.pop(session_id)
+            session = self._sessions[session_id]
             await session.kill()
+            self._sessions.pop(session_id, None)
 
     async def _spawn(
         self,

tests/tools/test_exec_session_tools.py

@@ -702,6 +702,30 @@ def test_terminate_by_owner_retains_failed_sessions():
     asyncio.run(run())
 
 
+def test_stale_cleanup_retains_session_when_kill_fails():
+    async def run() -> None:
+        manager = ExecSessionManager(idle_timeout=1)
+        session = SimpleNamespace(
+            session_id="stale-failed",
+            owner_session_key="cli:a",
+            last_access=time.monotonic() - 10,
+            kill=AsyncMock(side_effect=OSError("termination failed")),
+        )
+        manager._sessions[session.session_id] = session
+
+        with pytest.raises(OSError, match="termination failed"):
+            await manager.list(owner_session_key="cli:a")
+
+        assert manager._sessions == {session.session_id: session}
+        session.kill.assert_awaited_once()
+
+        session.kill.side_effect = None
+        assert await manager.list(owner_session_key="cli:a") == []
+        assert manager._sessions == {}
+
+    asyncio.run(run())
+
+
 def test_terminate_by_owner_skips_sessions_without_owner_key(tmp_path):
     async def run() -> None:
         manager = ExecSessionManager()