[OpenAI Fixes Codex Quota Anomaly Caused by Cache Hit Rate Issue, Resets All Account Limits] According to monitoring by Beating, OpenAI Codex team engineering lead Thibault Sottiaux announced that they have identified and fixed an underlying defect that caused the Codex intelligent programming system to consume usage quotas abnormally quickly, and have reset the quota limits for all affected accounts. Sottiaux pointed out that the abnormal quota consumption stemmed from a previously deployed performance optimization measure. During the rollback of the optimization, the team discovered a significant drop in cache hit rate during long-session context compression. When the system attempted to compress long session histories to free up space, the cache failed to hit properly, forcing Codex to recalculate and process the entire historical context, leading to an abnormal depletion of token quotas. OpenAI has now removed the optimization code that caused the issue, and the system has returned to normal. To compensate users affected by the system defect, the official team has reset the Codex usage quotas for all accounts. [Original Link]