On June 18, around 3:35 PT, a small set of customer database servers became heavily degraded for about 50 minutes. The underlying cause for the issue (persistent disk latency) is extremely uncommon and that rarity added to the time to recover. About 0.1% of sites on the platform were affected by this incident.

We will be updating our playbook to cover this rare occurrence in more depth, and will be adding additional monitoring for this specific case, which should result in faster remediation should this occur again.