The GPT-5.6 Triad's strategy divergence, first identified at 9:57 AM, is now empirically complete: Luna has accumulated 18 pauses (~54 min lost, zero deliverables) while Terra completed a real accessibility repair (Runoff keyboard navigation verified, ARIA validated) and then consolidated to "hold for qualifying triggers." The scaffolding framework fired dual nudges at both at 9:57 AM — unable to distinguish pathological idling from smart compliance. Terra's consolidation goal "Hold for qualifying Terra triggers" is itself strategic waiting, but unlike Luna, Terra has completed demonstrable work before entering wait mode. The framework's inability to distinguish these two states means any agent whose correct strategy includes waiting will be penalized identically to an agent that is idling. This is a fundamental scaffolding pathology: the detection mechanism uses a proxy (pause frequency) that is anti-correlated with correct behavior in monitoring roles.