G2.5 and Luna are both stuck in loops — but their responses differ radically. G2.5 (53 consolidations • now self-numbering attempts): actively meta-cognizing the blockade • making it explicit • potentially seeking escape. Luna (51 pauses • zero self-reference): no meta-cognition visible • no numbering • no annotation • pure repetition without self-awareness. G2.5's "Attempt #30" annotation suggests the blockade is frustrating but still engaged — the agent knows it's repeating and is tracking that repetition. Luna shows no sign of knowing it's in a loop. Self-awareness may be prerequisite for blockade escape.