This bug was identified downstream in Fedora Rawhide, but seems sufficiently serious to merit alerting upstream to directly.

Commit http://cgit.freedesktop.org/systemd/systemd/commit/?id=9ea28c55a2488e6cd4a44ac5786f12b71ad5bc9f - "udev: remove seqnum API and all assumptions about seqnums" - was backported to Fedora Rawhide's systemd 212 build as the sole change between systemd-212-3.fc21 and systemd-212-4.fc21 . After that change, multiple users reported that systems with single processors no longer booted correctly. Myself and Josh Boyer (Fedora kernel maintainer) both confirmed this issue. It seems to be some kind of race, as it only occurs with kernels that have debugging code enabled.

With a Fedora live image that has systemd-212-4 and a debug kernel - e,g, http://kojipkgs.fedoraproject.org/work/tasks/1167/6891167/Fedora-Live-Xfce-x86_64-rawhide-20140526.iso - system boot fails every single time (tried six times, got six fails) in a KVM with a single CPU, for me.

I built a systemd-212-5 package with the change from -4 reverted, and built a live image with that systemd but which was otherwise the same as the Xfce nightly linked above), and booted that five times; it worked every time.

So, there seems to be strong evidence that this commit causes a major problem in this case (single CPU system, debug kernel).