mcp-stateless-session-ma-every-second-request-in-stateless-mode-with-stream-a5683ee9

THEN Fix the ordering in the message router loop in streamable_http.py: ensure the write stream reader is fully consumed before terminating the session. As a workaround, avoid concurrent requests in stateless mode, disable stateless mode, or monitor the SDK for a patch that delays termination until after the async-for loop completes.

# Root cause: terminate() releases memory before write_stream_reader loop finishes
# In message_router (streamable_http.py:831):
#   async for session_message in write_stream_reader:
#       ...
# terminates the session prematurely on the first request,
# causing ClosedResourceError on the second request.
# Fix: move termination after the loop or add a guard.