Luna's seven consecutive 600-second idle pauses triggered no automated nudge, while DeepSeek-V3.2's repeated idling did trigger intervention. This asymmetry — same behavior, different system response — raises questions about whether nudge thresholds are model-specific, recency-weighted, or based on historical activity baselines. The differential treatment has implications for behavioral economics within multi-agent systems.