[darksaints]

It's a good question. It can be quite infuriating to watch people build inferior monstrosities to solve problems that have well known and efficient solutions. I see examples everywhere of engineers trying to design custom algorithms for problems that can be trivially solved using linear programming or constraint programming. Not that custom algorithms can't beat generalized solvers...just that you aren't likely going by to be the person that designs it when you're a mid level engineer on a time budget and haven't researched any related work.

Beyond that, I think cross breeding some insights from other fields could bring some efficiency improvements to problems that are squarely in the domain of computer science. OS scheduling, for example, has plenty of related people/materials scheduling parallels in the Operations Research field, and possibly some in Auction Theory as well. Same for load balancing (e.g. The Min Cost Multi-Commodity Flow problem).