Not really a program, but this year the Tabu [0] heuristic saved me at least on that order of time. I work on combinatorial optimization problem (think traveling salesman), typically we use local search heuristics like simulated annealing or parallel tempering that work well but require a lot of hyper parameter tuning. Tabu is super simple, has very few parameters to tune, and works great.
How do you implement "tabuList.contains(sCandidate)"?
Has it been successfully generalized to problems with infinite / continuous search spaces? I imagine you could use edit distance or something for large combinatorial spaces.
[0] https://en.wikipedia.org/wiki/Tabu_search