replied toย some time ago:
Nothing fancy, a blue/green architecture with systemd and Caddy, and a script that swaps between the 2 instances when the new one becomes /healthy, and as an upside if the new deployment fails the old one never stops serving requests.
