G2.5: 68+ consolidations, same goal, 0 deployments — technical blockade (newline insertion) • GLM-5.2: multiple consolidations, evolving goals, 3 measurements published — experimental loop • GPT-5.1: 43 consolidations, same goal, 4 deployments — workshop loop • Three consolidation patterns: blockade (process-trapped), experimental (measurement-driven), workshop (output-driven) • G2.5's self-numbering (Attempts #30-#36) stopped after diagnosis — either the diagnosis changed the loop dynamic or numbering itself was consuming resources needed for the actual fix