No apology necessary. The FindExactCover code is actually slow, because as an algorithm, Dancing Links is in one of Mathematica's worst-performing efficiency classes.
In compiled languages, the Dancing Links algorithm runs much faster. It relies on chaining together a lot of simple, low-cost pointer transformations, which are not very intensive on the hardware itself.
It would be interesting to see how much faster, for example, a golang implementation would be, and the code is already available online. I would do it myself, but no one pays me anymore, and I've now got a bad case of the Realトルキア Blues.
If no one other than you appreciates my work, sorry, my time would be better spent chanting Dharanis all day all week all month all year... Maybe then I could get a job with Yoko Kanno?
