On June 16, a bunch of Codex users got slapped with a "model at capacity" error, briefly taking the service offline.
https://twitter.com/thsottiaux/status/2066956441173323943
OpenAI's core product lead, Thibault Sottiaux, copped to the bug on X and said the team was on it. He later confirmed the fix was in, and that OpenAI would reset Codex rate limits for all subscription plans within 24 hours.