Increasing the new txs size/share limit is sticking your head in the sand about attack vectors. Unless of course your use case is a private pool aided by other hashers.
I do not consider making all users of p2pool receive and transmit up to 1 MB of data every 30 seconds (on average, assuming the attacker has 100% of the hashrate) to be an attack vector. Veqtrus keeps repeating that claim as if 1 MB is a lot of data. It's not.
If you have at least 4 Mbps of upstream and downstream bandwidth available that can be used by p2pool during an adversarial attack, then you can easily survive the "DoS attack" that veqtrus is so worried about. 4 Mbps is enough to receive a new 1MB share in about 2 seconds, in theory. It should be enough to allow you to keep up with the network okay if the share download time is 1/15th of the average share interval.
I consider 4 Mbps to be a reasonable minimum HW spec for p2pool, especially if it only applies to adversarial conditions with an attacker controlling nearly 100% of the hashrate. I really don't get why veqtrus is making such a big fuss about it.
For reference, in non-adversarial conditions, my p2pool nodes on jtoomimnet are each currently using around 13.3 KB/s (0.1 Mbps) in each direction, averaged over the last week.