### Check for existing issues

- [x] I have searched the existing issues and checked that my issue is not a duplicate. - Closely related (keys, closed as stale): #13210 - "ResetBudgetJob loads entire tables into memory -> OOM" - This report is specifically about the **team** reset path, which has the same defect and is still present in `v1.89.4`.

## What happened?

We set **team-level budgets on a large number of teams (20k+ roughly) all with the same `budget_duration` / reset duration of `7d`**.

Because every team shares the same reset duration, all of their `budget_reset_at` timestamps cluster together, so they all become "due for reset" within the same scheduler tick. When the `ResetBudgetJob` fires:

- Proxy process RSS spikes dramatically in a single reset tick. - The pod is **OOMKilled** and **restarts**. - The `reset_budget_job` scheduler job **fails to complete** (and after restart, the same set of teams is still due, so the crash loop repeats on the next tick).

**What I expected to happen:** The scheduled team budget reset should run without loading the entire set of expired teams into memory at once. It should page/stream through the teams in bounded chunks so mem…