Thanks from me too Barry, that's a good link
@jeff, in the screenshot that you showed, they could just be using a really high resolution normal map for each character, instead of a low-res one plus a microbump map.
At one games studio, I spent a week implementing a fancy skin shader that allowed the artists to blend between 4 different micro-bump textures over different parts of the skin... but in the end the texture artist told me that he didn't want to use it, and would rather just be allowed to author 2048x2048 resolution normal maps.
Micro-bump is bascially just sampling a second normal map (tiled repeatedly) and "adding" it somehow to your main normal map. There's lots of ways to combine the two, which greatly affects the results.
Last time I implemented it, I used the "Partial Derivative Normal Maps" technique from the link below, which allows you to add two "partial derivative normal maps" together and then renormalize them, and get a sensible normal as a result:
http://www.insomniacgames.com/tech/articles/1108/files/Ratchet_and_Clank_WWS_Debrief_Feb_08.pdf
These aren't about microbump, but some older links on skin shading:
http://http.developer.nvidia.com/GPUGems3/gpugems3_ch14.html
http://http.developer.nvidia.com/GPUGems/gpugems_ch16.html