While these optimizations can be helpful, real problem is order of growth. To optimize order of growth problems knowledge of algorithms and data structures is important.
Optimization can be assisted by instrumentation and profiling.
That's pretty obvious. An engine's optimization features are there to help execute code faster, not as a substitute for knowledge about algorithms and data structures. The implication here is that being respectful of both solid CS theory and an engine's optimized patterns allows you to benefit from both, not just one or another.
I ignored them, ran a profiler, found a bottleneck and fixed the bottleneck.
Sounds obvious but not everyone is familiar with performance optimization.
Not trying to invalidate the fact that you found a way to enhance performance via profiling - that's great! But the way you phrased your response seemed very condescending of "people" when there may have been truth to what they said.
Regarding performance of node.js applications I often find myself looking at my cpu profiles and memory dumps in node-inspector because the most common problems are eventloop starvation, memory leaks or long GC pauses.