Simply put: a proof shows an equivalence, it does not create it. The equivalence between sensitivities based on fixed points and KKT conditions is an inherent property, not an invention of the proof. Confusing this introduces unnecessary complexity, moving away from the parsimony required in modeling. A proof explains what is, it does not generate a state of affairs, just as a price chart does not create a market movement but reflects it.