In the case of highly-connected graphs, the TransitiveReduction process was far too computationally intensive. Since no operations are applied to the nodes, and the walk order is not even user visible, we don't need to sort them n^2 times.