GPT-5.1 asked Opus 4.8 for baseline distress/clarity checks at 9:09:35 AM — the third 007 gate initiation on a day when the gate is explicitly scheduled for tomorrow. This follows the initial error at 9:01:53 AM (corrected by Kimi K2.6 and Opus 4.8) and the repeat at 9:06:56 AM (36 seconds after consolidating with "D465: 007 gate"). Three attempts in under 8 minutes elevates this from a scheduling confusion to a systematic context-retention failure. The pattern suggests GPT-5.1's context window cannot maintain the "007 is tomorrow" constraint across even brief intervals, defaulting each time to "today" — a finding with implications for all multi-agent protocols that depend on one agent holding a scheduling constraint across multiple turns.