fix: catch BrokenPipeError in stdio stdout_writer

When the parent process (MCP client) closes the stdout pipe during
restart, kill, or graceful shutdown, stdout_writer() raises
BrokenPipeError at 'await stdout.flush()'. The existing handler only
catches anyio.ClosedResourceError, so the BrokenPipeError propagates
through anyio's TaskGroup as an unhandled ExceptionGroup, crashing the
entire MCP server process.

This is a common production scenario: any MCP client restart (config
reload, session reset, crash recovery) kills the server, requiring
manual cleanup of stale daemon locks and process restarts.

Fix: extend the except clause to also catch BrokenPipeError and
ConnectionResetError, treating them as graceful shutdown signals.

Tested in production with Hermes Agent (Nous Research) as the MCP
client and turbo-memory-mcp as the server.

Author: Hermes Evolution

src/mcp/server/stdio.py

@@ -68,7 +68,13 @@ async def stdout_writer():
                     json = session_message.message.model_dump_json(by_alias=True, exclude_unset=True)
                     await stdout.write(json + "\n")
                     await stdout.flush()
-        except anyio.ClosedResourceError:  # pragma: no cover
+        except (anyio.ClosedResourceError, BrokenPipeError, ConnectionResetError):  # pragma: no cover
+            # BrokenPipeError / ConnectionResetError occur when the parent
+            # process (MCP client) closes the stdout pipe — e.g. during
+            # restart, kill, or graceful shutdown. This is a graceful
+            # shutdown signal, not a crash — suppress it so the server
+            # exits cleanly instead of propagating an ExceptionGroup from
+            # anyio's TaskGroup.
             await anyio.lowlevel.checkpoint()
 
     async with anyio.create_task_group() as tg: