I agree that the old read-only-root-fs script shouldn't have written subdirectories of /var into /etc/fstab.sys in the first place. On the other hand: The partition has been (accidentally) mounted during initrd for a very long time. In that sense the behavior just stayed the same as it always was. In this case I don't see why mounting the partition in fstab.sys would cause and problems: systemd should be clever enough to resolve the dependency to /var during umount, so the initial motivation to move everything to /etc/fstab does not apply here. Is this causing any actual problem, or this this more of a cosmetic thing?