Back in the 80's and 90's, working on consoles in very tight space constraints, there were times we’d go with a horribly inefficient implementation of something like sort, because we only needed to sort 8 things and the savings in bytes of code was more important than the savings in time from a “better” algorithm.

Those abominations could be tweaked throughout the development of the game, and along the way we'd sometimes "rediscover" a more common (sane) algorithm. The refinement in this paper reminded me a lot of that.