Windows Codex app freezes due to stuck Git subprocesses

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

26.715.2305.0

What subscription do you have?

Plus

What platform is your computer?

Windows

What issue are you seeing?

I am experiencing a reproducible freeze in the Codex section of the new ChatGPT desktop app for Windows.

Environment

Windows 10
ChatGPT/Codex app package version: 26.715.2305.0
Chromium runtime: 150.0.7871.124
Git version: 2.53.0
Executable:
C:\Program Files\WindowsApps\OpenAI.Codex_26.715.2305.0_x64__2p2nqsd0c76g0\app\ChatGPT.exe

Symptoms

The application freezes and Windows reports “Not Responding”.
Windows Event Viewer records Application Hang, Event ID 1002.
The freeze can occur before I manually open a project.
Reinstalling, repairing, resetting the app and testing with a new Windows user did not solve it.
Restarting Windows temporarily clears the issue.

What steps can reproduce the bug?

Observed Git behavior

The Codex sidebar appears to scan several repositories simultaneously using commands such as:

git rev-parse
git status
git ls-files
reading remote.origin

One larger repository has a 6.5 MB Git index, and a direct git status takes approximately 13.5 seconds.

Some Git processes finish, while others remain stuck. Codex logs a timeout after approximately 60 seconds but appears unable to terminate or correctly reap the subprocesses. The queries are then started again, accumulating Git processes and eventually blocking the Electron interface.

In one especially relevant case, git rev-parse produced the expected 26-byte repository path output, but Codex still waited 60 seconds and reported a timeout. This suggests that the Git process had already responded, but the Codex Git worker failed to detect process termination or became stuck handling the process pipes.

Factors already ruled out

No abandoned .lock files.
Git hooks and fsmonitor are disabled by the Codex commands.
No repository-specific Git configuration appears to cause the issue.
Git was not recently updated.
The same Git commands complete normally when run outside the Codex worker.
No NVIDIA driver failure was recorded during the incident.
safe.directory ownership errors return within approximately 0.1 seconds and do not explain the 60–90 second stuck processes.

Likely trigger

This appears to be a regression in Git subprocess management on Windows, amplified by the sidebar scanning multiple repositories, including older repositories and one larger repository.

The setting:

git-show-sidebar-pr-icons=false

does not prevent the issue because the scans now appear to originate from sidebar_workspace_groups.

A useful reproduction test may be to leave only one repository in the Codex sidebar and compare the behavior.

I can provide logs, screenshots, Event Viewer details and process command lines if needed.

What is the expected behavior?

_No response_

Additional information

_No response_