Upscaling is working as expected. We'll be writing up a full report of what happened and what we're doing about it, but here's a summary:
At 2025-08-07 00:25:00 UTC we ingested some unexpected queue metrics for a far-future date. It looks like a customer triggered this from their laptop. Our upscale logic looks at the most recent metrics bucket as a boundary, so this data broke our upscaling.
We have many automated checks in place, but none caught this, and unfortunately it happened overnig...