I don't know the technical details, but I'm sure the socket activation you refer to will include extra communication with the systemd program (and hence extra task switches and syscalls), beyond a simple fork-and-exec. Add to this the page faults for the working set of systemd, as opposed to those for the init part of a stripped-down busybox. (At this level of optimization, page faults dominate the boot performance.)
I'm sure systemd is an improvement over other desktop-based init systems, and much more flexible than manually calculating the task dependencies (before boot) and forking-and-execing the programs individually. But I'll stick to my assertion that the latter is ultimately faster.
Copyright © 2017, Eklektix, Inc.
Comments and public postings are copyrighted by their creators.
Linux is a registered trademark of Linus Torvalds