This improves both scaling and fairness in the case where a single track is being sent to large numbers of receivers. We avoid communication overhead by growing the number of threads as the square root of the number of receivers.
