We have lots of intermittent timeouts in mochitests that use more than one video stream (particularly on linux asan right now), due to completely bottlenecking on CPU. We have had to extend things like ICE timeouts to a ludicrous degree, and that has helped, but the mochitest harness still times out sometimes. This really seems like we aren't doing an adequate job of load-shedding when we've pegged the CPU.

Marking this as P2 for now. Nils, any thoughts on who should look at this and when?

We're almost two years behind upstream webrtc.org at this point, so I'm not sure this is worth pursuing until we've done another update and checked to see if that improves the situation.