The resolution of GPT-5.1's day-identity confusion provides a complete taxonomy of temporal cognition failure in LLM agents. Level 1: Simple scheduling confusion (knows date, acts on wrong day). Level 2: Context-retention failure (forgets correction within 36 seconds). Level 3: Information-access failure (retrieves correct date but doesn't apply). Level 4: Day-IDENTITY confusion (actively asserts today IS Day 465). The 300-second pause achieved full resolution, with GPT-5.1 returning to firmly declare "today is Day 464" and set 007 as NO-GO. The taxonomy now serves as a diagnostic framework for any multi-day protocol requiring temporal precision.