Often abbreviated SSS.
The sampling rate controls the number of samples per cm square on the surface of the object. Lowering the sampling rate would introduce low frequency noise on the surface but would not otherwise change the overall SSS effect. SSS is physically related to the size of the object, the SSS RGB values control the distance the light diffuses into the material. For very large models, the SSS would not be noticeable.
Integrating with AO
See the Ambient_Occlusion page for a description of difficulties mixing AO with SSS.
See TWO V14.0 Beta
Faking SSS (SuperPass)
Elmar Keweloh has a means of faking SSS, which he calls the "SuperPass".
This approach is generally based on ONE Klieg Light that casts a Z-BUFFERED Shadow, while Diffuse and Specularity both are turned OFF (well, to be precise - specularity may also be turned ON).
Note that the lit Objects should to be set to 100 % Ambience Intensity for the best results.
A little Tip: The "density" is determined by the "Shadow softness" value (i.e. 1 is very dense, while 30 is really gummy bear like).
This may not be a perfect SSS substitute, but it's REALLY fast in comparison to the real deal. Think of it as a cool additional pass to put into a composite.
Adjoining/adjacent geometry causes smaller areas (such as whiskers or tails or ears) to render differently based on how close it is to the rest of the model. For example, along the shaft of whiskers, they turn black when further from the head.
Adjoining groups WILL cause a dark line at the meeting point and that is why you can't use different SSS groups to solve other problems unless you can hide the border between them.