History-conditional assessment

Reusing assessment sequences to select, train, or adapt later checkpoints invalidates the ordinary independent-assessment hypothesis. Fresh noise on reused sequences controls a finite-design noise average, not automatically a fresh-sequence population target.