Hi,
that sounds really cool.
Interestingly I had very similar ideas while thinking about dags and also hoped at the beginning, two orderings would suffice for each and any graph, just to find out that it does not.
Did you further follow the idea with N orderings?