Update — Elevated response times (~60s)
We’ve been mitigating elevated response times (up to ~60 seconds) affecting some requests.

This was triggered by a very short burst where incoming traffic peaked at roughly 10× normal levels. Our autoscaling did not react quickly enough because scaling signals were not interpreted correctly under the spike conditions. Within ~30 seconds of the peak, we initiated a manual scale-up to stabilize capacity.

The updated autoscaling configuration is now in production. It improves how we interpret CPU, memory, and worker utilization, and it also adds request rate as an additional scaling trigger so we can respond faster to sudden traffic bursts.

We’re continuing to monitor closely to confirm sustained stability and will share another update if anything changes.

API recovered.

API went down.