Jheriko: why not add a synched 3D SSC ring at the intersections? Could help hide the edges and would give the 'door' a bit of depth.
Some optimizations tricks:
c=2*(sqr(ox)+sqr(oy)+sqr(oz)-1);
ox,oy,oz only shift once per sphere... i'm not sure about this, but if you can spare a couple of vars you could probably speed this up by calculating this once (for every sphere) per frame rather than per pixel.
d21s1=sqrt(sqr(ox+t1*x1-cx1)+sqr(oy+t1*y1-cy1)+sqr(oz+t1*z1-cz1));
d22s1=sqrt(sqr(ox+t2*x1-cx1)+sqr(oy+t2*y1-cy1)+sqr(oz+t2*z1-cz1));
t1=if(below(d21s1,1),100,if(below(t1,0),100,t1));
t2=if(below(d22s1,1),100,if(below(t2,0),100,t2));
You can lose the square roots here 😉. In general, when comparing distances, just compare the squares... here it doesn't matter because you're comparing with 1.
Oh and normally because of the way the quadratic equation for a sphere it built-up, you can always pick out one of the two solutions, because one will always be the backfacing point, the other the front-facing point.
Just make sure you save several copies of this preset (I noticed you have the same filename for both). I've lost a bunch of stuff like this. It never hurts to backup and save every 10 minutes. After every major change, I keep a 'last good' version before trying something else.
As far as realism goes, it's really up to you to make it look as convincing as possible: the raytracing is correct, there's nothing to change about that. One thing I would play with is fog. You could use exponential decay, inverse decay, inverse square decay (with a minor negative offset to guarantee zero visibility at a certain distance).