Finally spent the time to read this whole thread yesterday after bookmarking it when it first started! Lots of great stuff. 
I thought I’d have a go at contributing.
As has been touched on before in this thread, when you start trying to do things more realistically, the scale of your objects and the scene start to become more important. Or at least, keeping some kind of ‘scale connection’ between your geometry and the shaders and lights you use. I’ve created some renders in Maxwell to demonstrate this (though you could do the same in mental ray etc). Oh, and I realise this thread is quite a few years in span, but these days Maxwell can use the N, K fresnel equation for its materials and also load measured data in the form of ‘complex IOR’ files so you get the right response per wavelength of light etc. It has a library of around 275 measured substances. mental ray also has measured BRDF shaders I believe but they aren’t included with the host applications yet.
Anyway, back to scale and materials! Let’s start with absorbtion (or ‘attenuation’ or ‘falloff’) for things like glass:
As you can see, my statue (from a free Stanford collection) is 1 meter tall and is made of glass. It’s important to know that even in very transparent substances like glass, the light does die off as it moves through the object (I guess because of impurities and the micro structure itself etc). Changing how fast the light is absorbed inside changes the look of the material. As far as I know, common glass has an attenuation distance of about 30cm, so the third statue is the most correct… for this scale! If instead of changing the attenuation distance, I scaled my scene up or down, I would notice the same changing appearance because the light is travelling further or shorter distances through the same substance.
Scale also matters for SSS (subsurface scattering) effects:
(Please ignore the ugly artifacting in the first two statues… the model is from scan data and so there must be holes and geometry issues causing it.)
Here I’m only changing what in Maxwell is called the scattering coefficient, which basically means how dense the particles are inside the object. It doesn’t really matter what the number means exactly for this demo, just that denser looks more solid. The more particles there are inside the object, the less likely the light is to make it all the way through. You also may be more likely to think that the object is of a smaller scale towards the left, and bigger on the right.
For that material the surface colour is a yellowish-orange, and the particles inside are white. Also, you often have with SSS an option to bias how much of the light scatters ‘forwards’ through the object, or ‘backwards’ back towards the light source. In this case my bias (called assymetry in Maxwell) is set to 0 so it’s 50/50 (equal in all directions, diffuse).
One last example (sorry if the images are too big, you can drag them into a new tab to see the full size):
Here is a combination of the above two effects, in a way. I have kept the SSS coefficient the same, but changed the absorbtion as with the glass example. Once the absorbtion is really low, the light can only penetrate 0.4 cm into the object and so you don’t get much influence from the white particles inside. The light is dieing off before it gets back out. Again, changing the scale of the model would do this also.
In mental ray, the absorbtion on the MIA (or Arch & Design material in 3DS Max) is called Refraction Falloff under the Advanced Refraction rollout. In terms of SSS being accurately changed with the scene scale, I think that only works in mental ray when using the Misss Physical shader, which uses photons.
imagine the “blurriness” of reflection as a continuum from “soft” (diffuse) to “sharp” (specular). Nothing in nature is perfectly diffuse or perfectly specular, so we might say that all reflections are glossy.
