This reminds me somewhat of research by (I believe) Andrew Miller (though I fail to find a reference right now) from around 2012 or so, where he suggested a consensus mechanism that (by using DAG structure) has difficulty adjust not based on hashrate, but based on percentage of blocks not included in the main chain. So there would be a target percentage of wasted (“orphaned”) blocks, and if the real percentage is too low, the difficulty would decrease, and if it was too high, it would increase. The idea is that the percentage wasted blocks determines the advantage a colluding attacker has against honest miners, so if we accept a fixed small percentage, the block rate should be as high as possible while staying below that fixed percentage.
However, I think it suffers from the same problem pointed out above, that it incentivizes changing the network topology to colocate miners.