How did this guy test systemd exactly?

If other packages want their software supervised by systemd they can just drop static service files into /lib/systemd/system and/or dynamic service files into /etc/systemd/system. And then they can tell systemd to start/stop/whatever the service as they wish via simple D-Bus calls.

I'd rather wait for those "oddball" services to show up and become a problem before I'd think of shifting socket creation into any external binaries.

So far we didn't have any problems with the current listening code, and Apple built an entire OS on this idea, so I think unless somebody shows me a real-life problem with this i rather not spend too much time thinking about non-existing problems.