A backlog in the webhook dispatcher caused outbound deliveries to be delayed by up to 4 minutes for ~25 minutes. The queue was drained automatically and worker concurrency was increased.
Vector index rebuild on a primary partition caused P95 retrieval latency to climb from ~120ms to ~480ms for approximately 12 minutes. No data loss; success rate remained at 100%.
Planned PostgreSQL minor version upgrade on the primary cluster. Replicas served reads throughout; brief writer failover (~8 seconds) at 03:14 UTC. No customer impact reported.
No other incidents in the last 90 days.
Email updates when an incident is opened, updated, or resolved. No marketing.