Edit: Makes me wonder how much better p2pool would be if it was also in C++......
c/c++ are much better suited to the task than python. Python may be quick for development but p2pool has outgrown it. I toyed with the idea of rewriting p2pool in c ages ago but it's a veritable bootload of work. On the other hand, I may find a way to hook something like p2pool into my ckpool code or vice versa.
Maybe a re-write could also be an opportunity to tinker with the protocol to improve things. Not sure if these are totally practical but a couple of thoughts I had:
At the moment, as I understand it, there is a single difficulty for the sharechain. I was penciling out a potential alternative whereby each node attempts to target 10 shares in the PPNS sharechain. This would be big enough to minimize volatility but small enough for there to be room in the PPNS for more lower hashrate miners.
Fixing the impact of the 30 second share time causing difficulties with some ASIC miners would good too. Maybe one possibility is that the last 20 shares in the sharechain are not paid. So miners would work on the same blockheader for 10 minutes not updating it for each and every share found. Then when the 20th new share is found a new blockheader is created that incorporates them and the process starts again. I think this would balance out over time to be fair and maybe fix some of the miner compatibility issues?
I was thinking about having shot at implementing p2pool in Go. But it's a big job and complicated. I don't think I have the time or skill to do it.