This line of play leads to this position:
[White "White"] [Black "Black to move"] [FEN "8/rp4p1/1N1nkb2/3n4/6R1/1R6/8/2K5 b - - 1 38"]
StockFish 10+ assures me that Black has a big advantage, but I am interested to know whether it is possible to humanly analyze it and be completely certain of a winning strategy. Although SF suggests keeping my pieces, my instinct is to exchange as many pieces as possible except the pawns, while pushing the pawns to queen, which includes taking the knight immediately. However, is there any hope of proving a winning strategy here, rather than relying on intuition? I think this is somewhat plausible because Black has enough minor pieces for the king to hide behind (even after exchanging 2 of them), and so White cannot make sustained checks. The question is whether there is a systematic way for Black to push the pawns that is guaranteed to work.
I do not see any convincing reason that there cannot be a proof of a winning strategy, because there are numerous examples of endgames with human analysis of the optimal strategy (e.g. a,b,c,d,e). If the position was not so imbalanced, I know it would be much harder to prove a win. But here, Black has sufficiently many pieces that I feel there should be a proof that we can force some trades to get to a certain winning position. That is also why I would choose to take the knight immediately to hopefully make a full analysis humanly tractable.