Just for point of discussion: why not two models? One featuring a high-poly mesh for closeups, and a second, lower-poly mesh with the details baked onto UV-mapped textures for shots that don't require precision?
That is certainly an option and one often used in CG production pipelines. My involvement is limited to building the "hero" model, from which Dennis'
F/X guys should be able to derive a lower-poly version if they find it advantageous. However, I am still hoping to strike a balance on the overall polycount, with added detail in the areas that need it for close-ups, that will work for the vast majority of shots.