Simulation Fidelity and Ontological Collapse: A Structural Paradox — LessWrong