I would love to see a clearer description of how the pools work once we're on mainnet. For now, it seems like pools should keep splitting delegates so that they always are in top 100, but with as few votes as possible, such as to maximize their number of eligible delegates. It's a bit of a race to the bottom.
Why not scale delegate rewards logarithmically with votes? That way higher-vote delegates get some benefit and incentive, but not perfectly scaling so they don't become overly powerful.
Interesting idea to do it logarithmically. I've be tempted to leave it alone though and just see what happens.
I think the system would be more fair though if the payouts to voters were handled automatically by the semux code itself versus relying on validators to write their own external code to split the rewards. Also, in that case, you wouldn't have to worry about some validators keeping all the money for themselves and not sharing it with voters (which most seem to do now except for at least two pools of validators). I suppose though the block size might get really, really big and full with reward splitting if too many voters were participating. If that happened every 30 seconds, it might just get too large.
If the block reward were split automatically by the semux code itself, it could do things like not charging a transaction fee for transactions related specifically to the payouts to voters.
If the block size would be too large paying every 30 seconds, the semux code could still manage payouts and do a once-a-day payout, but it would have to keep track of data of all the data being accumulated throughout the day in the database. The code would be trickier, but have much smaller overall block usage.