> having proper distributed service management
Which is more or less what CoreOS's `fleet` already does on top of systemd. They used a different protocol for remote communications (etcd, ie. HTTP+JSON) vs. local, which I guess can be a sensible choice as requirements are quite different.