OpenAI formed an internal "war room" after users reported that its AI coding agent, Codex, was exhausting usage credits faster than expected. The Codex team identified several issues, including overly proactive auto-reviews, excessive sub-agent activity, duplicate background suggestions, and repeated retries. The company has since deployed fixes and granted all users an additional usage reset.