Well, it depends on what you mean by "override."
A winning bitcoin block is sent to the bitcoin network and to the p2pool network. Bitcoin will payout the reward normally for the found block (including the finder's share). However, if the winning block isn't on the longest chain of the p2pool sharechain, it will get orphaned by the p2pool network.
In that case, the winning block is paid out and the solver will get credit for it for the current block (in addition to the 0.5% bonus for finding it), but the person that solved it won't get a credit for the share it for future blocks on the p2pool chain. This is fair because everyone has the same chance of it happening, and it does not mathematically affect expected payout for anyone. Additionally, this is extremely rare (<0.1% of blocks found will have this).
Ironically this just happened to me on the last block.
holy crap:
https://blockchain.info/tx/270d0e76ac7bbe95720b95651a427bb75112005c4be8e57e3be14dadffc6ea9dwhat are the chances of that!?

This is perfectly the example of how p2pool doesn't waste more work than solo mining! This counted as a p2pool "reject" but it still paid out!
I will admit that my math was wrong, it's > 0.1% of blocks. It should be ~ the DOA/stale rate, which is pretty high actually. See your stats:
"Note that blocks may have been orphaned from the P2Pool chain and so not be here."