One edge POP returned 502 for ~12% of EU traffic during a scheduled provider window. Fastly rerouted automatically; we updated the failover threshold so it triggers ~6 min earlier next time.
Background re-index job stalled after a database connection pool exhaustion. Pool size doubled; alert threshold lowered from 90% → 75%.
Stripe webhook retry storm during their planned migration. No payments were lost; affected webhooks replayed automatically.