It would seem at first glance that the extra round trip(s) would be less costly than the potencial bottleneck by a large margin. As mentioned several times in

It would seem at first glance that the extra round trip(s) would be less costly than the potencial bottleneck by a large margin. As mentioned several times in the other thread, increasing average latency to narrow your latency histogram is almost always the correct choice.