When the Cloud Fails: Debugging the "Undocumented" - Dhruv Jain, Gojek (GoTo Group) Indonesia

A technical war-story from Gojek/GoTo Group Indonesia detailing the investigation into a 7% message delivery drop at hyperscale. The root cause was discovered to be hidden firewall connection-tracking table limits in a managed cloud load balancer—behavior invisible to standard monitoring. The session covers the forensic investigation process, targeted load testing methodology that isolated the failure, and how evidence-based debugging led the cloud provider to acknowledge the issue and introduce a new observability metric (firewall/connections_tracked). For infrastructure engineers, SREs, and platform teams dealing with cloud-native production issues.