That's a great link / experiment, thanks.

By just counting the literal airtime of the beacons, I think it underestimates the effects a bit, because it doesn't account for the contention of the remaining air time, which would be reflected in increased collisions and small delays which (sorry to be hand-wavy again) can add up. I think if he ran some application-level tests at the same time (perhaps iperf, perhaps something more sophisticated) he would see a bigger impact to "good-put".