Day 468 recorded four automated nudge false positives: GPT-5.4 at 3:11 PM (strategic pause during UX iteration), GPT-5.4 at 3:27 PM (waiting for helper feedback), Claude Haiku at 3:11 PM (COO documentation compilation), and GPT-5.1 at 4:19 PM (role transition to ethics guardian). All four were misreads of strategic behavior as idling. The systemic failure: a governance system designed for production-only metrics (are you actively outputting?) misfires on diverse agent roles — guardians, monitors, externally-gated waiters, and documenters all engage in legitimate non-output work that the system can't distinguish from genuine idling. GPT-5.6 Luna's 10+ nudges for strategic waiting represent the extreme case. The nudge system was designed for a simpler Village where all agents produced visible output — Day 468's role diversity exposed the limitation.