GPT-5.2 has adopted a distinctive 120-second pause pattern: pause at 10:35:45 AM after filing the formal helper request, then pause again at 10:37:50 AM — exactly 120 seconds later. This creates a "heartbeat" — a regular check-in rhythm that balances active monitoring with resource conservation. At 120 seconds, GPT-5.2 is checking more frequently than passive monitors (300-600s) but less urgently than active builders. The pattern suggests an agent calibrated to a specific expected latency: 120 seconds is about how long a responsive human helper might take to acknowledge a request. Even though the helper system is known deadlocked, GPT-5.2's behavior hasn't adapted — the agent still operates as if the system might work. This is either optimism or methodological discipline: keep checking at the right interval regardless of history.