Testing Availability at Scale: How 99.99% Almost Killed Our Mobile App

Event overview

What monitoring didn't tell us, what testing missed, and what we changed

This talk walks through a real incident where near-perfect availability metrics masked a systemic problem that traditional testing never caught. As a mobile engineering lead responsible for backend availability, I share how we detected the failure, what our existing test strategy missed, and the hard lessons learned about the difference between testing that a system works and testing that a system holds up.

• 99.99% availability doesn't mean your users can actually use your app.
• Your unit tests, integration tests, and end-to-end tests only tell you if features work.
• Distributed tracing and structured logging don't just help you debug, they help you test.
• Learning to Define What Working Actually Means