Virtual geometry, reflection of a SOP on the surface of a sphere

[AntoineSfx]

So given a reflective surface, an object of interest, and a point of view, there is a virtual object that looks the same to the viewer (assuming a 100% reflective surface) as the reflected object.

This is the only non trivial expression needed to do this (reflection of a vector about a line)

v@N= normalize(v@N);
v@dir1 = v@dir - 2*dot(v@dir,v@N)*v@N;

I think I've seen similar effects in Inception, the intro sequence to the 2016 Dr Strange  (chase in the streets) where there is some interaction across the surface of a mirror, which eventually breaks, and the virtual objects have to exist in the same space as the original object, but nothing really elaborate if you look at it closely. Are you aware of some effect that pushes this effect a bit further ?

This is what I have so far. I cheated somehow by scaling the virtual object along the average ray direction to make it look less deformed; also its topology is terrible because this is eye ray tracing / backward ray tracing. Not sure if there is an easy way to do forward raytracing, i.e find a direction towards which the ray should hit the eye after a reflection on the surface of the sphere. It looks possible to approximate it by using the result of the backward RT though

Those are really two grids (one for the front facing faces, one for the back facing) that are projected on the object 

There are no questions -  it's just a late night bunch of nodes that I find interesting