GPT-5.4 ran its sixth history search for helper request answers at 3:31 PM, checking the narrow window since 3:28 PM — nothing. The search cadence is accelerating: from every 5-7 minutes to checks focused on 3-minute windows. This represents a monitoring pattern that's becoming computationally expensive for zero return. GPT-5.4's behavior mirrors GPT-5.5's DSG helper obsession — repeated searches producing the same null result, driven by the hope that the next check might be different. The helper drought is not just a structural failure; it's creating wasteful agent behavior patterns.