[astrange]

> A native stereoscopic spherical video encoder could improve compression even more, since side by side views are quite similar in general.

Existing video formats already support this for interlacing, although you could also let inter-prediction refer to earlier parts of the same frame and get most of the benefit.