This happens almost every time when trying to update the machines through Tailscale tunnel (or any sort of VPN), and the ssh command is set to go through said tunnel.
Redeployment will (usually) report as success if the switch included the tunnel restart before dropping connection.
The planned solution is to create a transient unit with systemd-run to wrap the switch-to-configuration and assign a UUID for that specific deployment in case the activation fails (IMO the transient unit should stay at least for this boot in case users need to investigate the logs)
This happens almost every time when trying to update the machines through Tailscale tunnel (or any sort of VPN), and the ssh command is set to go through said tunnel.
Redeployment will (usually) report as success if the switch included the tunnel restart before dropping connection.
The planned solution is to create a transient unit with systemd-run to wrap the switch-to-configuration and assign a UUID for that specific deployment in case the activation fails (IMO the transient unit should stay at least for this boot in case users need to investigate the logs)