Flaky CI test: test_run_git_timeout races the timeout against a fast git command

[agentforce314]

What fails

tests/test_git_utilities.py::TestRunGit::test_run_git_timeout

assert rc != 0 or stdout == ""
AssertionError: assert (0 != 0 or 'commit 41bc6...' == '')

Why it's flaky

The test runs a real git command under a small timeout and asserts the timeout won: either a non-zero rc or empty stdout. On a fast/cold-cache CI runner the git command completes before the timeout fires, returning rc=0 with a full commit log — so the assertion fails without any code being wrong. It's a race between a wall-clock timeout and command latency.

Evidence it's environmental, not code

• Failed on fix/tui-agent-urls-clickable (PR fix(ui-tui): render agent URLs as visible, clickable text #694, run 29178945090) — a diff containing zero Python changes.
• Failed identically on dependabot/pip/rich-gte-13.9.4 (run 28982750863, 2026-07-08), then passed on re-run of the same branch the next day.
• Passes locally 3/3 consecutive runs (macOS, Python 3.11.15).

Suggested fix

Make the command deterministically outlast the timeout instead of racing a real git log, e.g. run git hash-object --stdin with stdin held open (blocks forever, timeout always triggers), or monkeypatch the subprocess to sleep. That converts the probabilistic assertion into a deterministic one.

Found by gstack /ship while landing PR #694.

🤖 Generated with Claude Code