I've seen a design like this before in, oh I don't know, every init system under the sun I think it was.
Incorrect. SysV init (and even more so, BSD) are basically a one-shot "run a bunch of things" (which amounts to a shell script that runs other shell scripts). Once booted, init does next to nothing. 99.999999% of it's life is spent monitoring the exit of a bunch getty's. It's not eating all logs (journalD). It's not managing networking. It's not passing messages around. In fact, it's listening for exactly ONE message: change runlevel, which only root can send.
Things like upstart and systemD are problems looking for problems. I ran linux as my desktop for years, LONG before Pottering learned to spell systemD. It worked just fucking fine. And I didn't have to constantly login as root to start and stop things. If I wanted a local webserver on the machine, it was set to start at boot. I didn't need systemD to sit there listening on port 80 waiting to start apache. (a function -- starting network services on demand, for the record, that was quite adequately served by inetd) In fact, I ran entire labs of linux desktops where the users could not login as root.