fix cancelled + serialize error

[mikasenghaas]

Description

• CancelledError no longer silently drops requests: Previously, if a process_request task was cancelled (e.g. client disconnect, task cancellation), the server returned without sending any ZMQ response — leaving the client hanging until the 10-hour default timeout. Now sends an error response so the client slot is freed immediately.
• Serialization failures send fallback error response: If msgpack.packb(response.model_dump()) throws (e.g. on a malformed or huge response), the exception was unhandled and socket.send_multipart was never reached — again leaving the client hanging. Now catches serialization errors and sends a plain error response.

Type of Change

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Documentation update
• Test improvement

Testing

• All existing tests pass when running uv run pytest locally.
• New tests have been added to cover the changes

Checklist

• My code follows the style guidelines of this project as outlined in AGENTS.md
• I have performed a self-review of my own code
• I have commented my code, particularly in hard-to-understand areas
• I have made corresponding changes to the documentation
• My changes generate no new warnings
• Any dependent changes have been merged and published

Additional Notes

---

Note

Medium Risk
Changes error-handling in the ZMQ request/response path to avoid dropping replies; a mistake could alter client-visible behavior or mask underlying failures, but scope is limited to response generation.

Overview
Prevents ZMQ clients from hanging when a process_request task is cancelled by returning a BaseResponse error instead of silently exiting.

Hardens response sending by catching exceptions during msgpack serialization and emitting a fallback error payload, ensuring send_multipart still runs and the client receives a failure response.

^{Written by Cursor Bugbot for commit d26f327. This will update automatically on new commits. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

[cursor]

Caught CancelledError re-raises at subsequent await points

High Severity

Catching asyncio.CancelledError at line 249 without calling asyncio.current_task().uncancel() leaves the task's internal cancellation flag set. The asyncio event loop will re-raise CancelledError at the next await point (await asyncio.to_thread(...) on line 272). Since CancelledError is a BaseException (not Exception) in Python 3.10+, the except Exception on line 273 won't catch it. The error propagates uncaught and the response is never sent — the exact problem this PR aims to fix.

Additional Locations (1)

• verifiers/workers/server/zmq_env_server.py#L270-L273