It seems like it should always be optimal to take the hill as fast as possible. What is "possible" is just a bit complicated, I suppose. It presume it depends on the remainder of the course, because if you have just one short hill left to the finish then you can utterly knacker yourself on it, but if you've another 100k to do after then you need to be more careful.

Ultimately I suspect this is in some way analogous to the various proposed refinements to simple speed-to-fly theory, which attempt to take into account the fact that one's altitude is finite. I have yet to read up on these in detail, but I gather the executive summary is just the intuitive "get less picky about lift as you get lower".