Yep, easy enough to workaround. Seems like a needless limitation however. Would be nice if someone on node.js core (or v8 team at google) would try to fix that bug.
Also wonder what other bugs are going to emerge as a result of trying to use a client side app on the server.
Do note clearly that V8's garbage collector is fairly slow (think old Java GC) so having more instances of node with smaller heaps is a good idea anyways.
Argh, I was afraid of that. Can we have something that is a hybrid of refcounting and stop-the-world GC? I read some nice academic paper a few years ago about how people made a GC that combined the benefits of refcounting and stop the world.
Depending on your workload, 1 node per 2 cores is a better rule of thumb. Also, split the networking across multiple NICs.