How we tracked down a 16-year-old SQLite bug
Summary
Tailscale details a 16-year-old SQLite WAL-Reset bug that caused cross-shard database corruption and service downtime. The post explains the investigative process, including live telemetry, a transaction-logging pipeline, and collaboration with SQLite maintainers. It ends with a fix and improved backup/recovery practices, underscoring the risks of non-standard configurations for boring tech like SQLite.