Question on this point. Do we really need a switch? At least in my case when I used Oculus CV and now HP Reverb, there is always a 2D pancake screen running simultaneously with VR. So technically we would not need to switch. 2D is always there. Would be a matter to see if the quality settings of that 2D auxiliary screen can be customized etc.
As far as I remember, the "mirror" screen is the view of the ?left? eye. So it would be a bit distorted and the view wouldn't entirely line-up with 'forwards'? (I seem to remember various shader effects would also not work properly)
The ships were all modelled in 3D so that was no problem, so if the avatar is modelled in 3D and is mapped and sent across the network as such, why should this be an issue (i know it's a big if, and I know little about this stuff).
Your ship is mapped to a single point in space e.g. 0,0,0. It has a direction and a bunch of constants/variables associated with it, but for the most part it is a fixed body (it doesn't bend/warp), so it's "just" tracking the motion of that single point.
In non-VR an avatar is pretty similar. It's based on a single point in space. A client deduces where the arms/legs are supposed to be, given current/past actions, the terrain and the avatars movements using a bunch of rules (IK) and canned animations.
In VR, it gets far more complicated. An avatar is based on ~3 points (head+left hand+right hand), trying to deduce the position of the skeleton and the current motion is significantly more intensive/problematic/buggy. (every client would need to derive the skeleton position for every frame - so quite a bit of
additional data would need to be sent and processed, on top of the load generated by the rest of the game).
It's probably not an insurmountable problem, other games do this - but EDs networking is unusual, and this doesn't feel like a particularly trivial thing to do. And then there's interacting with the environment (where a lot of players might expect something HL:Alyx-style).