Obviously trying to do an optimization designed to reduce the impact of latency between client <-> server is going to have no impact if you are testing on localhost where latency is already effectively zero.

That's not to say i think this optimization is neccesarily worth it, just that testing on localhost is not a good test of this.