Global resets can nullify banked-reset value with only hours of notice, undermining quota trust for heavy users

Summary

This is not a compensation request. It is a product-policy / quota-design issue that materially damages trust for heavy Codex users.

Banked resets are presented as something users should use strategically. But a server-side global reset can arrive with only a few hours of notice and effectively erase most of the practical value of a recently redeemed banked reset.

That creates the opposite of a reward: users become afraid to redeem a reset because a surprise global reset may make that decision look wasteful shortly afterward.

Concrete case

• Subscription: high-usage paid Codex plan
• Banked reset redeemed: 2026-09-06 17:00 KST (UTC+9)
• A global reset was then applied around 2026-09-08 ~10:00 KST
• Immediately before the global reset, the weekly allowance was still around 90% remaining
• After the global reset, it became 100%

So the global reset provided only about 10% incremental benefit, while the previously redeemed banked reset had already been consumed.

The problem is not that the global reset happened. The problem is that users are asked to treat banked resets as scarce strategic resources, while OpenAI can later overwrite the same quota state with little practical advance notice.

For users in Korea and other non-US time zones, an announcement only a few hours before a global reset may happen during normal sleeping hours, making it impossible to react even if they follow official channels.

Why this is a product problem

This creates a bad incentive structure:

1. Banked resets are scarce and have meaningful practical value for heavy users.
2. Users are encouraged to decide carefully when to redeem them.
3. Surprise / short-notice global resets can retroactively make that decision wasteful.
4. The rational user behavior becomes: do not trust the product UI; monitor social media and community channels before redeeming a reset.

That should not be the optimal strategy for a paid product.

A reward system should reduce friction and increase goodwill. In the current design, the reward can instead create anxiety and regret.

Requested product changes

Please treat this as a quota reliability / product trust issue, not as a goodwill or refund issue.

Possible fixes:

1. Give planned global resets meaningful advance notice (for example 24–48 hours), in-product, not only on social media.
2. Protect recently redeemed banked resets when a global reset occurs before their new weekly window has meaningfully run its course. The exact mechanism could be re-crediting the reset, preserving remaining value, or another equivalent protection.
3. Show reset event history in the Usage UI: event type, timestamp, trigger, and whether a banked reset was consumed.
4. Document the interaction between banked resets and global/admin resets so users can make decisions without guessing.
5. If a global reset is intentionally meant as a reward, ensure it does not create a worse outcome for users who happened to redeem a scarce reset shortly beforehand.

Retention impact

For heavy users, predictability of quota is part of the product quality.

If a user cannot trust that a strategically redeemed reset will retain its value for a reasonable period, then the model can be excellent and the overall product can still feel unreliable.

This is the kind of issue that can make a paying power user reconsider staying with ChatGPT/Codex and move to a competitor with more predictable usage rules.

Again: I am not asking for a personal replacement reset or compensation. I am asking for the system to be fixed so this does not keep happening to users.

No repository files, tests, or entry points are named. Start by determining whether this repository owns the quota and reset behavior, then narrow the alternatives to one agreed policy. Done requires a specific, testable rule for banked resets and global resets, plus clear notice or usage-history requirements.