I'm not sure about Edge in particular, but on most platform runtimes "timers" are separate from IO-availability-triggered asynchronous events. When a WebRTC data connection signals its send channel has become unblocked/receive channel has data available, a handler will be called right then; there's no timer involved. Timers are just for things that have no relevant IO to control them, like animations.
That was my understanding as well, but the improvement in CPU utilization they showed [0] seemed to be a bit too dramatic to be attributed solely to more infrequent timer ticks. I guess they must done some serious cherry-picking on the sites used for that benchmark.
