You can get a net positive utility gain from a net zero money gain with a continuous function, if it were say, logarithmic. A has 1X$ and B has 10X$, then an exchange occurs where 1X$ moves from B to A, now it's 2X$ and 9X$, but the change in happiness for A could be log(2/1)~=.30 while B log(9/10) ~=-.04

This is just one utility vs wealth model that would argue wealth transfers improve net utility in some contexts.