Hypergrowth isn't always easy
Summary
A Tailscale blog post acknowledges recent uptime challenges during rapid growth, emphasizes transparency via a public uptime history, and explains the architecture of the coordination service as a distributed message bus. It clarifies how outages primarily affect control-plane tasks while the data plane often remains operational, and outlines concrete improvements to reduce blast radius and downtime. The post also discusses future architecture work, including caching the network map, sharding, hot spares, and expanded multi-tailnet sharing.