Iteration 59 — H-060 EXCLUDED M4 ACTIVELY HARMFUL · DRO FAMILY 0-FOR-2

In plain words: the surviving cousin failed the last hurdle in the most damning way available. The question was: plugged into our verdict machinery on strictly held-out later regimes, do its readings help? They don't just fail to help — they actively make verdicts WORSE, by more than random pairing noise ever does. And because the machinery is free to use a reading upside-down if that helps, this can't be fixed by flipping the sign: the reading's relationship to the truth flips on its own between the training era and the test era — an instrument whose arrow points a different way each season. One more scientific deposit though: for the third time, through a third lens, features that look WORSE by a stability-adjacent measure turn out MORE likely to keep their edge — the inverted mechanism first seen in iteration 9 is now corroborated three ways. Both worst-case-risk cousins are now honorably dead, and the queue moves to the regime-variance candidate.