Windows desktop app: every new thread shows "Reconnecting... timeout waiting for child process to exit" during startup websocket prewarm

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

Codex Desktop package: OpenAI.Codex_26.506.3484.0_x64__2p2nqsd0c76g0

What subscription do you have?

gpt-pro 5x

What platform is your computer?

OS: Windows 11

What issue are you seeing?

Environment

• OS: Windows 11
• Codex Desktop package: OpenAI.Codex_26.506.3484.0_x64__2p2nqsd0c76g0
• Codex app version from logs: 0.129.0-alpha.15
• Auth mode: ChatGPT
• Model: gpt-5.5
• Network: user proxy enabled at 127.0.0.1:10808
• WinHTTP proxy was initially direct, later synced to 127.0.0.1:10808, but issue persists
• Browser plugin disabled, issue persists

Problem

On Windows Codex Desktop, every new conversation/thread shows reconnect attempts on the first message:

Reconnecting... 2/5
timeout waiting for child process to exit

After retries, the response usually continues. Existing/ warmed threads work better. The issue appears thread-startup specific.

Logs

The local logs show:

codex_core::session_startup_prewarm
startup websocket prewarm setup failed: timeout waiting for child process to exit

startup websocket prewarm timed out before the first turn could use it


### What steps can reproduce the bug?

when starting the Codex desktop, it will happen when sending the first message to AI for each session.

### What is the expected behavior?

Starting a new thread should not trigger reconnect attempts or child process timeout.

### Additional information

What I tried
Restarted Codex Desktop
Disabled browser-use/browser plugin
Synced WinHTTP proxy with user proxy:
netsh winhttp import proxy source=ie
Confirmed WinHTTP proxy:
127.0.0.1:10808
Removed custom large context config:
model_context_window = 1000000
model_auto_compact_token_limit = 900000