It occurred to me that a way to test the given explanation that the speed difference is due to the effect of the buffering on TCP traffic would be to measure the total bandwidth used by multiple downloads or multiple types of traffic running at the same time. If the given explanation is correct, I would expect multiple downloads or multiple types of traffic to be able to utilise more bandwidth in total than a single download.
I decided to have 10 ping commands running, each one sending and receiving a 1500 byte packet per second, which in total uses up about 15KB/s of bandwidth up and down, shown in the first image. PN and BT profiles both correct at 3.3 Mb. wget reports the speed of a single download running at about 380KB/s. When the ping commands run at the same time as the download, the download speed drops to about 365-368KB/s. The third and fourth images shown when the ping commands start and finish, there's a visible increase in upstream bandwidth used, but there's no increase in total downstream bandwidth used.
