Eric Chen comments on On expected utility, part 3: VNM, separability, and more