@empty[g]
Pooling will be de-incentivized with PoCW because miners with very low hashrate can survive solo mining, there will be no reason for a small miner with few S9s to give up with his sovereignty by joining a pool and paying fees for this at the same time.
As of selfish mining by big farms, I've discussed it up-thread with @anonymint but I haven't checked whether after his ban and the removal of his posts and the ones which has quoted them, the selfish mining related discussions are still comprehensive enough or not. So let's briefly take a look at this problem once more:
First of all, selfish mining in its essence should not be categorized as a flaw even in legacy PoW.
A miner with any power can choose to keep his mined blocks private (and putting them in the risk of becoming orphan) hoping to find next block(s). Traditional winner-takes-all approach in bitcoin correctly assumes that for miners such a trade off won't be encouraging enough but many authors have tried to find a justification for miners with a large hash rate share (> 30%) to take advantage of selfish mining.
Although
Sirer et all classical analysis suggests feasibility of such an attack
* on bitcoin and we have now at least one pool with such a large share (BTC.com) and historically Slush pool reached to even higher thresholds temporarily, there is no sign of a selfish mining attack on bitcoin blockchain, ever.
Such an attack would cause an unexpected increase in orphan rate, trivially detectable. If Sirer is correct(which I doubt it), it should be probably because a large pool/farm with up to 30% network share has long term interests in keeping system secure rather than seeking a few percent more profits.
Now that we have a more precise understanding of this threat in PoW, it is time to do a comparative analysis for PoCW:
1- PoCW de-incentivizes joining pools, so selfish mining should be committed by a farm rather than a pool. But a mining farm with 30%+ network hash power is many times harder to be established compared to a pool.
2- A pool operator has less risk factors than a farm owner involved. A pool operator is gambling with his 1-2% shares but a farm owner is risking total block's reward.
3- In PoW, winner-takes-all implies both large risks and large rewards to be gambled, in PoCW miners should decide about every single share they generate gradually during each round. It would be easily provable that there is no advantage in keeping shares private in the first stages of contribution phase and it would yield little advantage in the latest stages because the selfish miner will have almost nothing more than a 3% reward incentive to start finalization phase earlier.
The 3rd argument above is of much importance and you should carefully consider the protocol details to grab it:
Suppose the network is already converged to a prepared block and its Merkle tree consequently. At first when shares are starting to accumulate, it is in the best interests of both regular and selfish miners to publish a share because no matter who finalizes the block, it would be included with a high probability.
As the shares accumulate more and more, in a critical threshold (90%+ ), a selfish miner with very large hashrate, may start to consider the possibility of keeping his (frequently generating) shares private to be more profitable.
Just like the case with PoW, a trade off should be made between the risk of losing both privately kept shares and the 3% reward of generating a finalized block on one hand and guaranteeing one's own shares to be included ( as there is always a chance for a share no to be included because a miner who manages to finalize a block puts his shares in prority) and gaining the 3% on the other hand.
I think an exact mathematical analysis is needed here but let's assume a miner with a very large share would find it more profitable to go selfish in the last stages, still the network would be like 95% contributive because the selfish miner needs to include most of the shares to prove his finalized block.
It is worth mentioning here that I have made an improvement in the protocol that allows Prepared blocks to freely use any of the n (n not being so large) older blocks as their parent for other reasons but one side effect of this improvement would be neutralizing any effort to keep Finalized blocks private and deceiving network to generate stall shares. It is why I deliberately ignored this scenario, but even without such an improvement a same analysis would supportively show how better is the situation with PoCW compared to traditional PoW.
*I have doubts in Sirer's work's integrity, because from a pure mathematical point of view it implies that a long range attack on bitcoin network is possible by less than 50% network share which is false because a far more rigorous mathematical analysis would show you need 50%+1 power for such an attack. If Sirer's analysis was correct, a selfish miner with 30% power would be able to keep his chain private for more than just few blocks and show up with a longer alternative chain which is practically impossible with just 30% share of the network's total power.