Your Postgres service was running and writing data throughout that hour: its logs show regular checkpoints between 12:36 and 13:36 GMT-3, writing about as much as in the periods before and after, so the database was not idle.

The flatline and the single spike look like a gap in metrics collection rather than real behavior. The point at 13:36 shows 72.8 vCPU against your 24 vCPU limit, which the service cannot actually use, so that point is consistent with the samples from the missing hour being recorded together in one bucket. The traffic in that hour happened as normal, it was just reported late. We haven't established why collection paused for that window.