HippyDrome, your site is very interestingā¦
The topology of your control mesh is pretty clean and ordered - itās actually very similar looking to a bunch of NURBS surfaces patched together. So it seems that you only make use of one advantage of subdivs; that it is one single surface without heavy mathematics to keep it in one piece. But you give up the other very important advantage and that is to localize detail (however you can do it with hierarchical editing). This implies that when you want to add more detail into an area, youāll end up adding vertices to other areas as well. For example the elbow and knee areas of your mesh are very dense; and you still donāt have too much definition there. I can see your reasoning behind this, that you most likely want to have a very smooth derived surface; I sometimes end up reworking my meshes for the same reasons. But I wouldnāt give up on localized detail for many reasons; like, my machine is quite slow 
A few areas on your model that could use more localized detail are for example the eyebrow (it has detail, but itās continued in the skull area too) and the corners of the eyes on the face; and the nee/elbow areas lack definition as Iāve mentioned above. All IMHO of course.
Also, the size of your datasets can become pretty important as well, more vertices mean larger scene files and longer render times. Iāve just seen a character yesterday that had a 200 MB RIB file (no mistake, 200 megs for one character!) - had it been built with such a ācleanā mesh, it wouldāve taken at least two times as much space. And all this data has to bee sent to the renderfarm machines on a network, many many times⦠
By the way your skinning looks pretty nice and clean. Can we please get some insight to your methods? Iām about to skin 5 human characters in Maya and we can say that Iām really scared about it⦠