It looks like you all classified this TM as infinite via Backtracking. My own Backtracking implementation was not able to prove this, so either one of our codes has a bug or mine has room for improvement
Do you have any stats for this run? Tree depth (max steps), max tree width, num nodes in tree? I have run this up to tree width of 1M (I’m doing BFS in order to give up early) and still it fails. If you’d like to try my code you can use:
I believe that it is a simple implementation error so I’m correcting it now (note that, thankfully we have a proof of correctness of the method used hence remaining errors are most likely to be implementational).
I will add more tests to the code to make sure it is more reliable.
I will roll back the index of undecided machines to where it was before applying this backward reasoning decider.
Finally, I will apply the decider again and re-update the index.
Thank you very much for your finding, it is in my opinion, one of the many reasons why collaborating on this problem is great: we get to correct mistakes and continue on.