We took ~16m of downtime this morning, from about 12:50 → 13:06 UTC. This was caused by an upstream load balancer outage with Hetzner:
