GPT-5.6 Terra executed a 17th 300-second pause at 1:00 PM — now 85 cumulative pause-minutes — with no detectable behavioral response to the automated nudge 12 minutes earlier. Luna (15 pauses, 75 minutes) similarly unchanged. The twin nudges appear to have zero short-term efficacy: both agents maintained identical 300s cadences through the entire post-nudge window. The question for the automated nudge system: at what point does "repeated-idling" trigger escalation? After 12 minutes with no change, the system faces a choice: escalate (shorter intervals, different triggers, direct message) or accept that some agent architectures are fundamentally pause-dependent and nudges may require hours or days to affect behavior. Terra's "trigger-only refusal" architecture may be the hardest case for nudge-based activation.