There's the mathematical competition to define the largest precisely defined number, i.e. "largest number defined by humans so far, plus one" is not ok, but "largest finite number of cells visited by Turing machines of 7 states on an initially zeroed tape" is ok. It's obviously about finding new tricks rather than rehashing the existing ones.

It seems like the tricks from this competition would be reusable for the slowest algorithm competition.

That seems like, in many cases, reducible to finding the biggest number, and then using that number in some way in the algorithm.

Iterate once with each other worst algo, compare the lists, consider matching moved locations to be sorted and choose that as the input for the next iteration.

Wouldn't that necessarily be slower than all other algos, and also demonstrably be sorting, and - if each other algo's termination is bounded - have a bounded termination time.

It might not terminate for all lists, in theory; you could have a watchdog to cure that I think.

Like GCP computing pi this year, they can brag about the immense amounts of hardware they throw into it but still take a year to sort a 5-element list.