If we wanted to further optimize this, I think it would mostly be a matter of a more performant OrderedSet implementation