Tailscale itself only uses sqlite[1], so I’m not sure if that really holds in this case.

[1]: https://tailscale.com/blog/database-for-2022

TIL! My problem with them requiring sqlite was that I assumed it would make a high availability setup either hard or impossible. Maybe that's not true, but definitely off the beaten path for headscale.

Headscale only supports a single control node.

I suppose there's always the old fashioned way of using drbd with heartbeat