Just a few comments:
First, I'll admit to not having all that much knowledge of Java. I've dabbled with it before, but wasn't all that amazed. Even so, I do know the general capabilities of the language and I'd place my bet on Java not being anything revolutionary. I can safely say this because Java doesn't incorperate anything revolutionary, language wise. If you hit the web and do some reading on languages and methods of advancing them, you'll see what it takes to become the next mass-language.
Java is an evolutionary language with language features that encourage saftey and ease of use. C++ is a language that tries to throw in the kitchen sink and lets the programmer sort it out. Each is best for different situations. Java is certainly powerful enough for a large project, and it contains many useful libraries that will certainly speed development times.
On the other hand, C++ has speed and some very useful features (does Java support templates?). Of course, the primary argument is that speed is only useful for 3D graphics, which isn't true. Extra speed allows you to do more with less, and as someone mentioned earlier, most people don't have the newest and fastest machines, yet the bar is constantly raised (I wish I could say I'm totally un-biased).
I personally prefer C++, as I've found it's a much more powerful language than most people think. It's also got many available libraries (such as SDL) that allow it to be quite cross-platform.
It all depends on the risk you wish to take, and what's best for the project, both merger and language choice considered. Most internet game projects don't last long, with death being slow, but well able to take out a large team just as well as a small one. Different perspectives don't always mesh as well as hoped. Clash is still going and OC3 is just starting. Is it worth the risk?
If it is, is dumping Java really the right way to go in terms of survival? Or should you chance it by switching to C++ and try to make something that pushes the boundries in terms of an internet project? The latter is certainly possible (check out
http://www.parsec.org for an example).
Just something to think about. While I think it'd be best to keep the projects separate, I'm sure you guys are more sure of your individual goals and compatibility, so are in a better position to make the right decision.
). 
