I'm pretty sure that paper is talking about an earlier version of this same work targeting Python instead of Ruby.
While it doesn't invalidate the reasoning behind the 'Pitfalls' paper, we think that we're addressing the problem in a fairly different manner: Instead of providing a fully featured JIT compiler that you need to wedge the language into, we've instead chosen to create a more modular system that lets the JIT compiler be more customized to the target language and VM.