By quickly looking at the source code, it seems to me that systemd did not really evolve on test coverage. That's a shame.
I use systemd everyday on my Arch Box-- its a desktop, VM host, home-server and media center. It is stable as can be aside from an issue thats come up in the last few days, which is a kernel issue, not a systemd problem. I say bring on systemd, bring on pulse, bring on networkmanager-- though NM needs bridging support ASAP to handle VM access.
Of course the coinciding thread quickly got down to the business of criticizing the neanderthals at Canonical for not embracing The One True Way (TM) that is, of course, whatever is proposed by either Lennart and / or Red Hat, and in this specific case, systemd.
The threads here are so reliably predictable and they're always good for a chuckle.
Can't wait for the next NVIDIA driver update article... followed by the cries of "still no KMS... still no Wayland support... still no Optimus... still not open source". Over and over again.
Yes, we get it... Upstart is not systemd. We shouldn't expect Canonical to bring systemd into Ubuntu any earlier than they announce that they're going to do so. Right now, Upstart is their way... and frankly, it works pretty damn well.
(And I love the snobbery of some here that UNIX systems are somehow less competent solutions because they don't have the awesomeness that is systemd or whatever other Linux fad-of-the-day.)
DeFacto open-source standard packages mean more development. Systemd has this. Upstart doesn't.
Systemd removes the absurdity of Dependencies & lets Linux really shine. This scales better to more CPUs (the future) while being great on 1 CPU too.
Scripting in standard startup is beta in nature. SystemD finalizes it in C, but leaves scripting open for one-off tasks.
SystemD (although having more purposes than Upstart) is nearer to the UNIX philosophy: "Do 1 thing & do it well" by ending the micromanagement once required of boot systems Upstart and Sys-V
Ubuntu's made great decisions, lets add to those by selecting the best init.
I doubt anyone here will listen to me, but what Canonical is doing in Ubuntu is actually better for system reliability. Developers capable of reviewing systemd on technical merits find that it makes Linux systems less reliable. I do not think that the systemd developers even attempt to dispute this. The issue is that the kernel will panic if PID 1 dies and systemd's design makes it very difficult to avoid failures in PID 1.