There is currently a regression issue with the video capture, so far confirmed only on HoloLens 2, where the Media Foundation capture engine successfully open the video capture device and acquires some video samples, but fails to write those video samples to its sink for a yet unknown reason.

Symptoms are, on the remote peer, that the video track coming from HoloLens 2 is negotiated and added (TrackAdded signal fired, optionally predated by RenegotiationNeeded if appropriate) but no data is sent over the network and consequently the per-frame callback I420RemoteVideoFrameReady is never fired, which e.g. for TestAppUWP will prevent the media player from starting.