Codex /responses timeouts on the ChatGPT subscription path (May 23-24) while status shows green - anyone else?

[aaajiao]

Setup: ChatGPT Codex subscription (OAuth), models gpt-5.5 and gpt-5.4-mini, driven by an agent gateway that calls https://chatgpt.com/backend-api/codex/responses.

Over the last ~24h I'm seeing frequent request timeouts on that endpoint:

• gpt-5.5 calls hit a 120 s fetch timeout (TimeoutError: request timed out / Request was aborted), repeatedly.
• Even gpt-5.4-mini (small, reasoning off) times out at a 15 s budget — a model that normally returns in 1–3 s.

status.openai.com currently shows Codex fully operational (no active incident), though history lists degraded latency on May 13 and an unavailable incident resolved May 23.

Questions for others on the subscription path:

1. Are you seeing elevated /responses latency / timeouts on gpt-5.x-codex around May 23–24, despite green status?
2. Is this a known under-reported backend degradation, or does it correlate with a plan tier / request shape?

Trying to distinguish a backend-side slowdown from something local. Concrete data points (model, latency, plan) appreciated.

[KP-2601]

yes as for 25 May, I'm still getting the time out, and it incredibly slower than it used to last week.

[ckcks12]

same here.
tried websocket, sse way.
both are slow.

detail :
both are ok to send requests but no responses
especially, no first byte at all.
so websocket or sse, whatever just retry on first byte timeout. (i set it like 2sec. tested 300+ calls with each call occurs with several tool calls and more than 5000 times hung, and first byte came in like wihtin 2s if successful)
without retry/timeout it tooks like 5min or something so what i received cloudflare timeout 529.
used gpt 5.5 thinking high, and here Asia Seoul and cf endpoint was seoul too.

claude ditched us, gpt overloaded. where to go -_-