The suite's cases, with their answers.
Per case, how many current models tested it and
got it fully right (the board's caseResults).
Now, for old drafts.
When the newest current test ran. A split case whose answer a person confirmed after it ("the answer is right") isn't listed again until a newer test splits on it.
What needs a look: easy, split, and duplicate cases, old drafts, and how the tags spread.
A suite's answer-key health, from its cases (retired ones are ignored) and how each case fared across the board's current tests. The floor of easy cases keeps the quick set's first, so the weekly check still catches a regression on the basics.