This started around August 25. After several days of troubleshooting on my own — reinstalls, sign-outs, network and device testing — I contacted Support on August 30.
I’m not looking for more troubleshooting suggestions—I’ve already tried everything Support requested, and they told me there is nothing else I need to do. I’m trying to understand whether other users have experienced this and, most importantly, how long it took to recover.
For over a week, my account has been seriously out of sync. The same cloud conversations show different content on the Web and Mac app. Prompts and answers are missing, reappear, or appear in the wrong order. I have also received “Error in message stream,” seen files disappear, and had a migration-created Project reappear after deletion.
At the same time, the model selector became unreliable. Chat sometimes appeared to route me to GPT-5.5 Mini, while Work kept reverting to Light instead of the higher-capability model I selected. The quality of the answers also dropped noticeably, with instructions repeatedly ignored. I don’t know whether this was caused by the synchronization problem, or a separate routing issue, but both started at the same time.
I reported everything on August 30. On August 31, OpenAI Support confirmed that the synchronization and content-consistency problem was still unresolved, but they could not give me an approximate timeline. They said no further action was needed from me at this time, and that they’d follow up through the case if there was meaningful progress or if they needed more information. It’s now been 5 business days with no update and no ETA, which has left me feeling like I’m stuck waiting with no way to escalate.
I can accept that some damaged chats or files may not be recoverable. What I need is for my account to synchronize normally again so that I can safely use the Web and Mac app, which I use for agentic work locally.
Has anyone experienced something similar?
Did your account eventually return to normal?
How long did it take?
Were new chats affected too?
Did normal model selection and answer quality return?
Did Support ever provide a meaningful update?
Thank you, guys.