The accounts that previously were best taking advantage of 0 minute self voting will receive a smaller % of the reward pool under HF20
Even if it does narrow the rewards gap over the current system, that doesn't make them the only two choices. In fact, I see no reason to think that what you're seeing has anything to do with "returning to the pool" - it's entirely a result of removing the extra author rewards, which I fully agree should be done.
When a vote "goes to the pool" - whether it's by this method, voting on a declined rewards post, flagging, or just sitting at 100% - it fundamentally broadens the rewards gap, because it's distributed proportionally by vote size. That effect here is swamped by the fact that you're removing a ton of rewards from certain large votes exploiting the bug.
Essentially you're taking a big chunk of money and narrowing the gap, and then using it to widen the gap again, differently, and somewhat less. Taking the money is probably a good idea, but it could be used in a different fashion that doesn't do that.
At least there you are finally saying that the net effect of hf20 is to narrow the gap.
I still think you are missing the big picture. The smallest accounts are always going to have the least amount of votes coming in shortly after posting and people jockying around for optimal curation reward vote time. The largest accounts are always going to have the largest amount of votes coming in shortly after posting and people jockying around for optimal curation reward vote time. If the early curation rewards are returned to the pool, the largest accounts will "send" proportionally more reward back to the pool - it decreases the % of the pool dominated by the large accounts.
If those rewards stay in the post pool, whether it is in the author reward for the post as in your proposal or the curation reward for the post as in earlier proposal advocated by @davemccoy among others, it doesn't do anything for the smaller accounts - the % of the pool dominated by the large accounts stays the same!
In effect this argument is that smaller accounts will be more effective at discouraging pre-fifteen-minute votes (relative to total votes received) than large accounts. That seems unlikely to me since larger accounts are almost universally more sophisticated users.
I've been pondering a project to vote accounts with large expected payouts at minute zero to create this effect intentionally. I think once more people are paying attention to this issue the consensus will be that it's incredibly stupid that I will be able to do that. But we will see.