I don’t see how this approach could work in an adversarial environment – latency isn’t a global constant, it’s something that varies depending on who you’re trying to connect with, and it’s something that you can optimise for simply by colocating.
If you initially have 100% of hashrate colocated in a single data centre with a latency of 1ms, and then introduce a new miner located on the opposite side of the world with half the original hashrate (so ending up with a 33%/67% split), and a latency of 300ms to the original data centre, how does the difficulty adjust to take this into account? (Presumably increasing from X at 1ms per block, to about 450*X at ~300ms per block with +50% hashrate?) Does this work if the original miner (with 67% hashrate) just ignores all the blocks from the new miner while the new miner’s tip is more than K (6? 100?) blocks’ less work?
How would this algorithm behave in the interplanetary scenario, where 50% of hashrate is on earth, 50% is on mars, and the latency between the two varies between 3min and 20min?
If it works, it seems very nifty.