Around 22:50 PDT, I tried jumping back to Primal from Crystal and was able to log in, so the housing timer has been addressed.

I am not yet convinced that this problem is resolved, given that, based on a tally-sheet, this was the 38th login attempt over the weekend, so I will do more validation testing this evening.

However, one thing that was clearly different was the length of the login queue at that point: while it was at least 20 during every failure, it was only 3 on success.

This would seem to make the idea of a database indexing problem the more likely cause, either because an inefficient lookup was able to succeed due to the reduced load at the time (and the subsequent updates may have corrected the error, at least for me), or because there's some background repair process that happened to run over my records between 9:30 PDT and 11:45 PDT.