Monthly Codex allowance can remain unreachable because of weekly caps, even after full resets

What feature would you like to see?

I would like the Codex usage limits to be aligned so that the monthly allowance can realistically be consumed within the monthly period.

My current Usage dashboard shows the following example (Sep 16, 2026):

• Monthly usage: 31 / 100 used (69% remaining), resets Oct 1
• 5-hour usage: 56% remaining
• Weekly usage: 55% remaining, resets Sep 22
• I have already used two Full / banked reset cards during this monthly period

The full resets gave me additional 5-hour / weekly capacity, but the monthly meter still has 69% remaining. At my current usage pattern, the shorter usage windows appear likely to become the practical bottleneck before I can get close to the monthly 100% limit.

This creates a confusing product model: the UI presents a monthly allowance, but stacked 5-hour and weekly ceilings can make a substantial part of that monthly allowance difficult or potentially impossible to use before the monthly reset, even for a heavy user who has already used additional full resets.

I am not asking to remove abuse/rate protections. A more balanced model could keep the 5-hour and weekly safeguards while allowing unused weekly allowance to roll forward within the same monthly period, capped by the monthly 100% allowance.

Possible approaches:

1. Allow unused weekly Codex allowance to roll over within the current monthly cycle, up to the monthly limit.
2. Size or dynamically adjust weekly limits so a consistently heavy user can reasonably reach the monthly allowance before it resets.
3. Clearly document how Monthly, Weekly, 5-hour, and Full/Banked Reset accounting interact.
4. In the Usage UI, show whether the remaining monthly allowance is actually reachable under the remaining weekly windows before the monthly reset.

Why this matters

For project-based development, usage is often bursty. A monthly allowance suggests that the user can consume that allowance during the month, while weekly windows can prevent that in practice. Full resets help temporarily, but if a user can use multiple full resets and still finish the month with a large unused monthly balance, the monthly meter becomes difficult to interpret as a meaningful usable allowance.

Additional information

There have been related discussions in #4217 and #19530, but the current Usage UI now explicitly exposes a Monthly limit alongside the 5-hour and Weekly limits. This report is specifically about the interaction of those three visible limits plus Full/Banked Resets in the current product.

I can provide a screenshot of the Usage dashboard showing the values above if helpful.

Start by reviewing related discussions in #4217 and #19530 and the current Usage dashboard behavior described here; no source file or test is named. A useful contribution would first pin down the intended interaction among Monthly, Weekly, 5-hour, and Full/Banked Reset limits, then identify the owning implementation and tests before changing behavior.