[macOS regression] list_threads hangs before AppServerConnection in delegated/background tasks

What version of the Codex App are you using (From “About Codex” dialog)?

26.814.41407

What subscription do you have?

Pro plan

What platform is your computer?

Darwin 25.6.0 arm64 arm

What issue are you seeing?

Agent-facing list_threads can remain pending indefinitely in delegated, agent-created, or other non-foreground tasks, before any corresponding local thread/list reaches AppServerConnection. On the same build it succeeds from a directly user-created foreground text task, while UI sidebar thread listing remains responsive.

What steps can reproduce the bug?

Feedback ID: 01a014bf-e57c-7bc2-8a73-0919e56207b2

1. Start a new voice chat in the Codex App.
2. Ask it to test the list_threads (“list chats”) tool with a supported limit.
3. The tool remains pending indefinitely and returns neither results nor an error.

The same call works from a directly user-created foreground text task. Rolling back from version 26.814.41407 to 26.803.61601 also fixes the issue.

What is the expected behavior?

list_threads should behave consistently across foreground and delegated/background task contexts, returning results or a bounded structured error.

Additional information

Verified:

• Failing agent calls produced no corresponding agent-originated local thread/list; UI-originated requests remained healthy.
• The same-build foreground text control returned normally.
• No SQLite lock/corruption or local app-server failure was found.
• Rolling back to 26.803.61601 fixes the issue.

This closely aligns with #37668 and seems related, but is not obviously identical. This reproduction adds a same-build foreground-vs-delegated control plus a working rollback version.

The evidence points to an execution-context/tool-routing disparity rather than pagination or thread-data corruption. Similar inconsistencies have also been observed in other non-user-initiated task flows.