Hey, I haven’t been around much, but I’ll give my two cents.
First, as time has gone on, I’ve soured on PoW in general due to articles like this (https://blog.sia.tech/the-state-of-cryptocurrency-mining-538004a37f9b). I’m not sure an ASIC free PoW is possible, and it seems big miner manufacturers are bad dudes (especially Bitmain). I fear that over time, if VRM becomes popular, we will end up in the same boat as all PoW coins, which seems bad.
Secondly, about the proposal. I think P2Pool is a failed experiment, so I don’t think implementing it is a good idea. I think the on-protocol pool is a good idea, so my vote would be for #3. However, I don’t know if a 10x decrease is good enough to avoid pools. In the current BTC mining environment even a 100x difficulty decrease would not be enough for a small ASIC farm to avoid having to use a pool.
I’ve been thinking about a solution to this problem: a hybrid PoW/PoS system, where blocks are created by PoW as usual, but rewards are distributed by a PoS-like mechanism.
People would mine special transactions as well as blocks. These special transactions (separate from the regular transactions) would act like shares of a pool (they have a difficulty that is less than the current block difficulty, but greater than some minimum amount) and these would be included in blocks.
The block reward would then be distributed to random share holders, in proportion to the amount of shares they have, and these shares would be consumed. Eventually someone mining a share would create a block, and they could broadcast it to the network.
If you set the minimum difficulty such that a small miner would get a share every day or so, you could have it, so people could use this to get rewards even if a very large number of miners were mining.
You might need a mechanism to prevent large miners from spamming the chain with shares, but I was thinking you could have a system where higher difficulty shares get proportionally more rewards, so big miners could put in a smaller number of high difficulty shares instead of a huge number of lower difficulty ones.
This might be too complex to implement right away, but a 10x difficulty reduction won’t be useful forever, so it might be a good idea to keep in mind.