Perhaps -- but even if that is true, interacting with more machines is presumably a trivially parallelizable operation, so it shouldn't double the overall respo

Perhaps -- but even if that is true, interacting with more machines is presumably a trivially parallelizable operation, so it shouldn't double the overall response latency.