I'd like to point out that by the time this article was posted, and thanks to some constructive discussion about the approach, the node-fib "project" on github

I'd like to point out that by the time this article was posted, and thanks to some constructive discussion about the approach, the node-fib "project" on github has been updated with a much faster recursive approach which doesn't use loads of memory and doesn't resort to memoisation, but still doesn't block the loop and serves almost as many requests/sec. My point here, and in writing the lib in the first place was that naive implementations are naive in any language/vm. I lump splitting an algorithm across the loop into the same band as deciding this task should spawn a thread/worker. And using child processes as workers is still an option in node.