We should add a small deterministic test for bicoloring, like @amontoison's rectangle. The issue is that for NaturalOrder and LargestFirst I get trivial colorings. Which order did you consider to make the nice pictures?