P2Pool actually has the opposite problem than most other pools. In your "typical" mining pool, the more hash rate the pool has, the smaller the effect of variance on the individual miner. In p2pool, the more hash rate the pool has, the larger the effect of variance on the individual miner. This is because of the nature of the share chain inherent in p2pool mining. For example, as of right now, p2pool has about 3PH/s on it. If you've got an Antminer S3, you can expect it to take over a day to find a single share to get you on the chain for payout.
I suspect that P2Pool will not be looked at as a "pool" for very much longer. As its hashrate grows, we are already starting to see "pools" built on top of P2Pool. Eventually, P2Pool will gain a significant share of the hashrate, but since miners win shares so infrequently, it will not be feasible for people to solo-mine against P2Pool anymore.
This will spur the creation of a "new P2Pool," which is a second distributed P2Pool built on top of the original P2Pool that uses the same exact technology but has more frequent block times. This is one way that faster confirmations can eventually be achieved without any fork at all, because it will be possible to see which transactions are being included in the "new P2Pool" and the original P2Pool's blocks. As a transaction gets included in a P2Pool up the chain, you can be more confident it will be included in the top-level blockchain.
That actually sounds pretty cool?
