Looks like its really a lot. Many of them will not "work" because body parts might intersect, however the order of magnitude should remain. I was thinking whether a "pose assistant" could be created, that accepts a rather unprecise description of the pose (e.g. place hand here and foot there and face target xyz) and that would then automatically find the pose with the least amount of overall bending to achieve whats asked. But the large number of possible poses seems to make that not an easy task.