[ChatGPT Web/Android][Sep 24–26, 2026] Frequent conversation load failures, 429 errors, and delayed image attachments #48393

Activity

1. github-actions

   added

   bugSomething isn't workingSomething isn't working

   connectivityIssues involving networking or endpoint connectivity problems (disconnections)Issues involving networking or endpoint connectivity problems (disconnections)

   rate-limitsIssues related to rate limits, quotas, and token usage reportingIssues related to rate limits, quotas, and token usage reporting

   on Sep 26, 2026
2. SIMON-WORLD commented on Sep 26, 2026

   

   SIMON-WORLD

   on Sep 26, 2026 – with ChatGPT Codex Connector

   More actions

   Additional severity update from ChatGPT Web on 2026-09-26:

   • ChatGPT Plus / Windows / ChatGPT Web.

   • At this observation point, nearly all existing conversations I tried to open failed with:

     Could not load this ChatGPT conversation

   • The conversation entries remained visible in the sidebar / Project UI; the failure occurred when opening the conversations.

   • This is a broader recurrence of my Sep 24 cross-surface reproduction in Cannot access chats #45259, where the same error was selective/intermittent and some existing conversations still loaded.

   • I did not perform additional diagnostic testing for this occurrence.

   Expected behavior: existing conversations that remain visible in the UI should load consistently.
3. tapperwijn89 commented on Sep 27, 2026

   

   tapperwijn89

   on Sep 27, 2026 – with ChatGPT Codex Connector

   More actions

   Confirming a matching web conversation-loading symptom from another Pro account on September 27, 2026.

   On chatgpt.com in Chrome/Windows, the contents of the same selected conversation appear and then disappear into the loading error. In an approximately 19-second recording this happens twice while the URL, sidebar and selected conversation remain present. During affected episodes all chats I try to open are affected, and keeping existing tabs open does not maintain access.

   I also see the "Too many requests" message about temporarily limiting conversation access. This continued from September 26 into the morning of September 27 and is preventing me from finishing website work.

   Full account of my observations: #48683. The recording does not show a network status code, and I am not claiming the Android or image-attachment symptoms from this report. Please acknowledge the continuing web impact and publish a mitigation or fix status.
4. lifeofgurpreet commented on Sep 29, 2026

   

   lifeofgurpreet

   on Sep 29, 2026 – with ChatGPT Codex Connector

   More actions

   Additional occurrence: ChatGPT web on macOS, September 30, 2026 — actual HTTP 429 responses captured

   I have lost access to existing ChatGPT conversations. Recent conversations remain visible in the sidebar, but selecting an existing conversation shows:

   Could not load this ChatGPT conversation
   Try again

   Two screenshots capture repeated HTTP 429 failures while this happens. The console contains failed conversation requests involving at least two distinct conversation IDs.

   Environment and time

   • Product: ChatGPT web at chatgpt.com.
   • Platform: macOS; the browser UI and developer tools appear to be Safari/WebKit. Exact OS and browser versions are not captured.
   • Screenshot times: 2026-09-30 10:01:03 and 10:01:55, UTC+11 — equivalent to 2026-09-29 23:01:03 and 23:01:55 UTC.
   • Subscription and model are not established by the screenshots.

   Observed interaction

   1. Open an existing conversation from Recents.
   2. The selected conversation remains listed in the sidebar.
   3. Its contents fail to load and the main pane displays the error above.
   4. The browser console shows the failed requests transcribed below.

   This describes the captured incident; the triggering conditions and reliable reproduction frequency are not yet known.

   Sanitized console evidence

   Repeated conversation-loading failures:

   Failed to load resource: the server responded with a status of 429 ()
   
   https://chatgpt.com/backend-api/conversations/[conversation-A]?num_turns=10&include_has_versions=true
   https://chatgpt.com/backend-api/conversation/[conversation-B]
   

   Additional errors visible during the same incident:

   WebSocket connection to wss://ws.chatgpt.com/[redacted] failed:
   WebSocket is closed before the connection is established.
   
   Failed to load resource: the server responded with a status of 403 ()
   
   https://chatgpt.com/api/codex-cloud/v1/environment-configs?scope=user&limit=10&omitDraft=true
   https://chatgpt.com/api/codex-cloud/v1/environment-configs?scope=workspace&limit=10&omitDraft=true
   

   The WebSocket and 403 errors are accompanying observations; their relationship to the conversation-loading failure is unconfirmed.

   Impact and expected behavior

   Existing conversation history is inaccessible, interrupting ongoing work and making the conversations appear lost. Existing chats should load reliably; a transient limit should provide clear retry guidance and recover cleanly.

   The screenshots establish failed conversation reads, but do not confirm permanent deletion. Response bodies, response headers, and a HAR were not captured in this evidence. Please investigate the continuing loss of conversation access and provide recovery or mitigation guidance.

   This adds macOS/Safari evidence with captured HTTP statuses to the conversation-loading symptoms in this issue and #48683. The raw screenshots contain account/conversation identifiers, unrelated chat titles, and a WebSocket verification value, so this public report uses a redacted text transcription.
5. mamamatison commented on Sep 30, 2026

   

   mamamatison

   on Sep 30, 2026

   More actions

   Adding a Sep 30 datapoint from another affected account, with browser traces.
   I’m seeing the same Web/Android failure family on Ubuntu/Chrome and the official Android app, using the same ChatGPT account.
   In multiple fresh web reproductions, including brand-new conversations, I captured:
   stall → /backend-api/f/conversation/resume 404 → repeated /backend-api/conversation/ 429 → UI network error → Retry temporarily recovers → later relapse
   In another brand-new chat, with DevTools already open before the failure, the Console first showed:
   WebSocket is closed before the connection is established
   followed by repeated 429s and a backend 404.
   The 429 response body was:
   {"detail":"Too many requests"}
   There was no Retry-After header.
   Example first-429 traces:
   Medium — 2026-09-30 10:28:56.825 CDT
   x-oai-request-id: 4803a5c6-4178-4971-9fc6-2b9f5c7996b6
   x-oai-turn-trace-id: a784efab-872d-43bb-981c-880d99fb7fa0
   High — 2026-09-30 10:29:06.825 CDT
   x-oai-request-id: 09812428-d965-43fb-b0e1-2008a1ed6a94
   x-oai-turn-trace-id: e1fabc98-cd61-47fa-9e3a-4f16910bc658
   Similar hanging/stalling was occurring concurrently in the Android app.
   OpenAI Support has escalated my account case. I have sanitized HAR captures covering the initial stall, 429/network-error state, Retry recovery, and post-Retry relapse.
   I’m not claiming a shared root cause without backend correlation, but this appears strongly related to the Web/Android conversation reliability failure family reported here.