[Windows] /feedback upload repeatedly fails and causes severe system slowdown
Open 💬 2 comments Opened Aug 12, 2026 by joon42066667-maker
💡 Likely answer: A maintainer (github-actions[bot], contributor)
responded on this thread — see the highlighted reply below.
Title: [Windows] /feedback upload repeatedly fails and causes severe system slowdown
I encountered a scheduled Automation failure in the ChatGPT/Codex desktop app and attempted to report it using /feedback.
The feedback upload repeatedly failed.
Environment:
- OS: Windows
- App: ChatGPT/Codex desktop app
- App version: Unknown
- Date: 2026-08-12 KST
Observed behavior:
- I opened
/feedbackfrom the existing Codex conversation containing the diagnostic report. - The first three upload attempts failed.
- During those attempts, the entire computer became severely unresponsive and the mouse cursor barely moved.
- The desktop app also terminated unexpectedly during an earlier attempt.
- I restarted both the app and the computer, but the problem continued.
- I then disabled “Include all browser tabs and browser logs” and tried again.
- The upload still failed with “Feedback upload failed.”
- The displayed feedback/session ID remained unchanged.
Expected behavior:
- The feedback should upload successfully.
- Collecting or uploading diagnostic information should not make the computer unresponsive or crash the app.
Underlying issue I was attempting to report:
- An ACTIVE weekday Automation scheduled for 2026-08-12 08:40 KST did not create a heartbeat Submission.
- The same Automation ran successfully the previous day.
- A separate Automation test succeeded later that day at 12:49:57 KST.
- A detailed diagnostic report is available.
- The relevant session/thread identifier is being kept privately and can be provided to an OpenAI team member if requested through an appropriate channel.
Attachments:
- Screenshot of the feedback upload failure
- Screenshot of the unexpected app termination
Sensitive identifiers have been removed from the public screenshots.
<img width="1960" height="1439" alt="Image" src="https://github.com/user-attachments/assets/12df5b2e-6ebb-4c92-8e5a-e5c417234251" />
<img width="1949" height="1436" alt="Image" src="https://github.com/user-attachments/assets/68803310-1aab-43c6-b67d-c59f9cca46a3" />
2 Comments
Potential duplicates detected. Please review them and close your issue if it is a duplicate.
Powered by Codex Action
Still reproducible on Windows Codex Desktop 26.820.60940.
I encountered a false-positive cyber-safety termination in a completely benign competitive-programming session (Codeforces 2251A). From that same affected session, I opened
/feedback, selected Safety check, and submitted with the current ChatGPT session logs enabled. The failing attempt also had the browser diagnostics option selected.The uploader returned:
Retrying does not complete the submission. The preserved client-side Feedback ID is:
The original safety false positive is documented in #40970. This creates a particularly bad failure chain: Codex terminates a substantial paid run through an apparent safety false positive, and then the built-in reporting path that is supposed to attach the affected session also fails. Because usage is limited, resuming or reproducing the interrupted task merely to report it would consume additional quota.
Expected behavior:
No private session logs, source code, local paths, or credentials are included in this public comment. A screenshot showing the exact uploader error and unchanged Feedback ID is available, but the connector used to post this comment cannot attach local binary files.