It sounds as though the experiment calculated all possible tetrahedra with the dataset. The null distribution was found by randomly generating similar datasets

It sounds as though the experiment calculated all possible tetrahedra with the dataset. The null distribution was found by randomly generating similar datasets (monte carlo). So it seems like you'd get the same result each time. The calculation sounds stupendously complicated. I can't wait to see it in next year's advent of code with some dbag claiming a sub-seconds run time.