Hmm... not much good using a fork designed to increase the max blocksize limit if you can't... increase the max blocksize.
It's a p2pool bug, not a Classic bug.
I see the issue was brought to forrestv's attention over 5 months ago on github - it would be nice if he followed it up or privided some additional info/update.....
Fixing this issue will likely require a hard fork of the p2pool share chain. (You would have to increase the max_remembered_txs_size value, which may result in your node creating shares that other nodes would not be able to accept or validate. I could be wrong though. This would be a good research project for someone who has time to read the code.)