No issues at all, and it is much faster with hardly any resource usage.
No need to compile it, it's ready to run from within the client folder. Yes, I run it with the standard p2pool code.
Well worth it

Edit: Makes me wonder how much better p2pool would be if it was also in C++......
I've also starting using the C++ client. It came down to RAM usage. Here's average usage on my node:
Matt's Java client: 150 MB
Matt's fork of p2pool + Relay Node: About 50 MB more RAM usage than vanilla p2pool
Matt's C++ client: 2MB
So since RAM is very limited on my node the C++ client is the clear winner. Also it gives you the flexibility to upgrade the Relay Node client without restarting p2pool which can be painful, mess up stats, and/or cost you shares.