Well I've done enough experiments to confirm a working hypothesis. It may not be 100% correct yet - there could be some exceptions. I did not do extensive experimenting with polygon density - just a handful of cases involving objects from 1 to a few hundred polygons.

Given a subject, such as a sphere in the middle of the scene, and you want to know how IBL or a self-lit surrounding prop such as the environment sphere or a box will affect the subject, here is the answer. IBL is effectively an infinite virtual sphere, so the analysis is the same as for a glowing prop that surrounds the scene, even if that prop isn't spherical.

Given any point on the subject, imagine a line passing from there to all the points of the glowing environment. If any given line passes through two opaque polygons, then the IDL contribution to lighting the subject from that area of the glowing environment is zero. Period. Doesn't matter which way these polygons are facing, or if shadows are off for the occluding polygons.

If there is nothing in between, the full effect of the glowing light source will be used.

If there is exactly ONE polygon between, then the amount of the glowing light source leaking through that polygon depends on the distance of that occluding polygon from the subject. If the distance is less than 1 inch, as I showed with props on the Ground or one-sided square as a wall, nearly 100% of the light will leak through. I tested this with a box, too, although it was difficult to get the camera inside a small box. If the distance to the occluding polygon is on the order of tens of feet, about half leaks through. If it is 750 feet (my EnvSphere radius), only about 6% leaks through. I could never get it to go to zero.

That's it.

Props that represent bounded volumes will reliably block IBL 100% because IBL will not leak through two polygons on the same line. In order to leak through a bounded volume, it has to go in and go out - that's two polygons. So a nearby box or cylinder always throws an occlusion shadow.

However, even a pair of parallel squares, perfectly in alignment, will block the light, although technically there is no inside/outside in this case. It's a simple matter of passing through two, even if simultaneously.

Any time you have only a single surface between the subject and the diffuse light source, the diffuse light will leak through. This is even true of nested spheres.

If you place two nested spheres around your subject, and you make them both black, your subject will receive no light from IDL. If the outer one is glowing red, the amount of red that leaks through will depend on the radius of the inner one, because the inner sphere is the only polygon between the subject and the outer sphere. The inner sphere glow will be used 100% in either case. The total diffuse luminance will be the inner sphere glow plus some fraction of the outer sphere glow.

If you have 3 nested spheres, the outermost, even if glowing, will not light the subject, because to reach it it would pass through two polygons.

I have demos, but I'm getting pretty tired. I'll post them another day.

---

Renderosity forum reply notifications are wonky. If I read a follow-up in a thread, but I don't myself reply, then notifications no longer happen AT ALL on that thread. So if I seem to be ignoring a question, that's why. (Updated September 23, 2019)