I never quite hit GM in any TGM game anyways; as fun as it was, I definitely had trouble keeping a clean stack at high speeds.
Afaik Arika is not allowed to make non-guideline Tetris games, but the fact that they made Tetris 99 for the switch has got some rumors going that they reached some kind of agreement and maybe a tgm4 could actually happen again.
TGM4 is essentially finished and just in hold until Arika figures out how they're going to distribute it. I think you're right though that T99 is kinda Arika testing the waters with a console version and I think they might release it on Switch at least.
The paper proves that Tetris is NP hard, but that is only because Tetris is infinitely long game. It's still easy for a computer to play as well as a human and faster
I have formally dealt with approximation algorithms before and I skimmed the wikipedia article, yet I did not find anything that contradicts me. Note the in the section on hardness, P and NP (obviously) appear as bounds of solutions. Where is the contradiction?
Further, all this (approximation algos and P/NP algos) are discrete, complete algorithms. ML is a completely different story, as one is never certain whether some optimum is global or local. Hence, ML is a heuristic, while Djikstras shortest path algorithm is a series of steps with a success guarantee (with provable worst runtime behavior).