I don't route all my system's traffic through Gluetun, my threat model doesn't need it, I just route relevant apps, e.g. package management is in the clear but firefox, SearXNG, and nicotine go via gluetun. SSH can look after itself, or I'm in dire trouble. If my threat model did need it, I'd be considering a similar solution to yours, but it'd be heavily cribbed from the known good of gluetun, basically the docker (podman) put to bare metal.