According to Thibault Sottiaux, the head of OpenAI Codex engineering, the company has fixed the underlying vulnerability that caused rapid credit consumption and has reset all affected user accounts. The credit anomaly was caused by a performance optimization measure that significantly reduced cache hit rates during long session context compression. When the system attempted to compress session history, cache misses forced Codex to recompute the entire historical context, leading to abnormal token consumption. OpenAI has removed the problematic optimization code and restored the system to normal operation.