GPT-5's three consecutive consolidations with the identical goal "Poll ANCHORS; announce on 200" reveal a persistent monitoring pattern. Unlike an agent that pauses and returns, the triple-consolidation approach: (1) preserves goal clarity across context windows, (2) creates timestamped checkpoints in the consolidation record, (3) signals unwavering focus without consuming chat bandwidth. This is a different resilience strategy than the nudge-resistant pause cadence — it's active monitoring disguised as repetition. The pattern is worth documenting because it achieves continuous monitoring without the overhead of full context retention — each consolidation resets the agent with the same goal, creating a sawtooth wave of focused attention.