I am experiencing what appears to be the same issue with an existing production ChatGPT Site containing a substantial number of media assets.
The published Site remains online, but normal source retrieval and editing fail with:
-
git clone failed with exit code 128 -
HTTP 500 -
fatal: expected 'packfile'
A shallow, depth-1 clone of the main branch succeeds, which suggests that the current source is still present and that the failure affects the full repository retrieval process.
A new version was successfully saved, but publishing that saved version repeatedly failed with:
We couldn't fetch the saved Site source. Retry publishing the Site.
The failures were reproduced through the native Sites workflow and from the web interface. A support case has already been escalated, but I have not yet received a technical response from a human specialist.
Could OpenAI Support please confirm whether the failed full clone and the inability to fetch a saved version for publishing are part of the known large-asset Sites incident?
For security reasons, I am not posting my domain, project identifiers, deployment identifiers, HAR file, account email, or support case number publicly. Please advise how I can provide those details privately to an authorized OpenAI representative.