Would p2pool have potential scalability issues if adopted on a scale much larger than now? For example, would the coinbase transaction be a problem with potentially tens of thousands of payout addresses? Also, if the solution to that issue is increasing p2pool share difficulty, isn't that just re-inventing something like solo mining in the long term?
You got it. You can only split a block so many ways. Or more correctly there is an upper bound on the number of p2pool shares per block because shares are added to a sharechain. That means there is a lower limit on share difficulty.
If there were no pools today and p2pool had 100% of the hashpower directed at it, then with 1 share per second the p2pool share difficulty 1/600th of the block difficulty and there would be on average 600 shares found per block.
It would be fascinating then if the centralized pools actually mined on top of p2pool themselves. P2pool could act as a sort of "granularity layer" dividing the block reward hundreds of ways. We would still have a mixture of centralized pools and maybe people mining on top of p2pool themselves directly, but the variance reduction incentive to join a large pool would decline dramatically.