I would not. And I agree, it will be worse than Haskell, which will only choke if the cpu intensive processes are all calculating things without allocating memory or doing IO (as opposed to node where IO is the only yield point).

And if the article talked about when Haskell green threads yield and compared that with node, as well as measured just how worse things are with a clustered setup, with several example problems, and by varying the parameters (such as number of processes)... then that would've been really awesome.