While discussing with my boss the potential for purchasing (or rather, recommending the purchasing of) some HPC equipment for one or more of our research groups, something occurred to me while I was making observations regarding the usage of their existing equipment.  Mike's Law of Academic Computoring is still in a pre-natal form, but it goes something like this:

The problem with things like high performance computing is that it is about getting things done.  Academia is not (necessarily) about getting things done, it is about getting as many papers published as quickly as possible.  If learning how to better use computing resources gets in the way of paper publication, then the learning takes a back seat.

Pretty long and gross, but it seems to sum up my experience with both procurement and "real-world" usage patterns so far.  As with anything there are, of course, exceptions.  But, in my experience, the academics who really want to continually learn how to use new technologies in order to fulfill their real ends are much harder to find than the ones who want things to never change so they can continue doing things in the ways they're used to.

Note that this is not (really) a value judgement - it doesn't matter to me how academics produce their work, only that they do.  But where academic departments run into trouble is when those who want New Ways conflict with those who feel more productive using the Old Ways, which will inevitably pinch the system administrators who have to support them.