close_workers hangs when closing already-gone or idle Codex sessions

[jalehman]

What

close_workers hangs indefinitely when trying to close Codex workers that have already completed or whose sessions are gone. The call never returns, requiring manual kill.

Reproduction

1. Spawn Codex workers via spawn_workers
2. Wait for them to complete (idle state)
3. Call close_workers with their session IDs
4. Command hangs — never returns

Expected

close_workers should return promptly, reporting which sessions were successfully closed and which were already gone/idle.

Observed

The MCP call blocks indefinitely. This has been consistently reproducible across multiple sessions. Workaround is to kill the mcporter process and let maniple's own cleanup handle worktree removal.

Context

This was noted in TOOLS.md for the OpenClaw main agent workspace as a known issue. Affects coordinator workflows where cleanup is the final step after cherry-picking worker commits.