BLAST.

14 chaos experiments scored across system tiers. 4 surfaced critical dependencies the runbook didn't name. 2 caused a full customer-facing outage during business hours.

You don't pick the day your dependencies fail. Unless you do.

CX-009 · DynamoDB partition failure PROD OUTAGE

Ran during business hours. Cart service down 47min. No fallback.

Add read replica failover, schedule chaos to off-hours.