vfs_mountroot() extracts the most recent mnt_time from the mounted filesystems (root and devfs at the time) and passes it to inittodr() as a fallback if there isn't a functioning system RTC. This is intended to ensure that things like log timestamps don't go backwards. Unfortunately, only the UFS code actually initialises mnt_time so other root filesystems types will leave mnt_time set to the epoch (1970-01-01 0000Z). Root on NFS isn't impacted because that uses nfs_mountroot(), which does inittodr() based on the remote filesystem. That leaves ZFS: If a system with root on ZFS doesn't have a functioning RTC then the system time will start at 1970-01-01 0000Z until (eg) NTP fixes it. This leaves close-to-epoch timestamps strewn throughout the filesystem, which is annoying.
Created a workaround for this issue on my RootOnZfs RPI4. https://www.freshports.org/sysutils/fakertc Putting myself in the cc of this PR.