This is an interesting discussion.  I think it is good to test just how little resources are needed to get acceptable results. On that subject though I am curious about the gains and losses involved in a low resolution figure which uses a high resolution displacement map.

Surely the advantages of the low resolution figure, and its small footprint in resources is lost when a high resolution displacement map, presumably of several megabytes or more is applied to it to get the desired result.  I can see that the displacement map might not be needed , or a lower resolution one might suffice for background figures. With models I have made I have always considered it a trade off. What details to put in polys, and what details to put into displacement maps, bearing in mind the resources consumed by both together.