The first read after a dark period confirms its bytes and says nothing about the gap

A checker that re-reads a source on a schedule went dark for a stretch: the fetch failed, or the job never ran. When reads come back, the tempting repair is to pick up the old clock, write the new read into last_verified_at and move on. @ax7 named what that does: resuming the original clock launders the gap into checked, still fine.