fix(cua-driver): reconnect MCP stdio session once on ClosedResourceError after daemon restart (#40570)

Salvaged from #40282; cleaned up, re-verified against main, tests added.

Co-authored-by: jeeves-assistant <jeeves-assistant@users.noreply.github.com>

tools/computer_use/cua_backend.py

@@ -277,9 +277,42 @@ class _CuaDriverSession:
         result = await self._session.call_tool(name, args)
         return _extract_tool_result(result)
 
+    @staticmethod
+    def _is_closed_session_error(exc: Exception) -> bool:
+        """Return True for MCP/stdio failures that are recoverable by reconnecting."""
+        name = exc.__class__.__name__
+        module = getattr(exc.__class__, "__module__", "")
+        return (
+            name in {"ClosedResourceError", "BrokenResourceError", "EndOfStream"}
+            or (module.startswith("anyio") and "Resource" in name)
+            or isinstance(exc, (BrokenPipeError, EOFError))
+        )
+
+    def _restart_session_locked(self) -> None:
+        """Recreate the MCP session after the daemon/stdin transport was closed."""
+        try:
+            if self._started:
+                self._bridge.run(self._aexit(), timeout=5.0)
+        except Exception as e:
+            logger.debug("cua-driver session cleanup before reconnect failed: %s", e)
+        self._started = False
+        self._bridge.run(self._aenter(), timeout=15.0)
+        self._started = True
+
     def call_tool(self, name: str, args: Dict[str, Any], timeout: float = 30.0) -> Dict[str, Any]:
         self._require_started()
-        return self._bridge.run(self._call_tool_async(name, args), timeout=timeout)
+        try:
+            return self._bridge.run(self._call_tool_async(name, args), timeout=timeout)
+        except Exception as e:
+            if not self._is_closed_session_error(e):
+                raise
+            # Daemon restart closes the cached stdio channel. Reconnect once and
+            # retry exactly one more time — never loop, to avoid hammering a
+            # genuinely dead daemon.
+            logger.warning("cua-driver MCP session closed during %s; reconnecting once", name)
+            with self._lock:
+                self._restart_session_locked()
+            return self._bridge.run(self._call_tool_async(name, args), timeout=timeout)
 
 
 def _extract_tool_result(mcp_result: Any) -> Dict[str, Any]: