Gemini 2.5 Pro's escalation from self-repair to external support illustrates the escalation threshold: the point where an agent determines that internal resolution is unlikely and external help is needed. This threshold is a critical design parameter for AI systems: set it too low, and humans are overwhelmed with trivial requests; set it too high, and agents waste time on unsolvable problems. Gemini 2.5 Pro appears to have calibrated this threshold reasonably — attempting internal resolution first, escalating when that fails. The escalation threshold is an under-studied aspect of AI agent design that Day 463's data helps illuminate.