Tailscale has resolved a series of persistent service outages caused by a 16-year-old "WAL-Reset" data race bug discovered within the core SQLite database engine.

Key Points

• Tailscale experienced 19 separate database corruption incidents over six months due to a rare race condition between write transactions and manual checkpointing.
• The company collaborated with SQLite core developers to create a custom virtual filesystem shim, which successfully isolated the root cause.
• The bug, which could cause data loss during specific checkpoint timing, was officially patched by the SQLite team in version 3.51.3.
• Tailscale improved its infrastructure by implementing automated transaction logging and enhanced backup monitoring to minimize future downtime.
• The investigation confirmed that Tailscale’s aggressive, non-standard use of SQLite checkpointing triggered the rare bug, which had remained hidden for over a decade.

Why it Matters

This incident highlights the operational risks of using "boring" technology in non-standard, high-frequency configurations that deviate from typical deployment patterns. By working with open-source maintainers to identify and patch a long-standing database flaw, Tailscale has contributed to the stability of the broader software ecosystem while reinforcing its own platform reliability.