OpenAI’s Codex engineering lead has driven repeated discretionary resets of weekly usage allowances for paid ChatGPT and Codex subscribers, typically to offset spikes from background features like auto-review, subagents, or retry logic, or after brief outages. Recent activity includes a global reset on September 26 following a service disruption, a banked reset on September 22 tied to GPT-6 model rollout, and earlier grants around the Astra launch, with announcements often confirming propagation within hours. These actions follow a recent pattern of roughly one reset every three days, well above the longer-term median, as the team iterates on token consumption and reporting accuracy. Trader sentiment reflects this cadence, tempered by the possibility of longer gaps during stable periods or major model transitions. Key upcoming catalysts include further GPT-6 refinements and any new usage-policy adjustments.