Testing Ferrari by seeing how much weight it can tow would actually get you pretty useful performance metric.
Point of this whole thing is that you cannot reasonably do complex computations in Node. While most "Web 2.0" applications probably does not need to do that in response to request, there are also many web applications that actually do something that might be reasonably called complex cpu-intensive processing in response to user request (although most such things are hard to scale). So it really boils down to "right tool for the job" - and for many practical (but maybe un-popular) problems, Node simply is not the right tool.
This is a serious question, because I honestly don't know the answer. What purpose does node serve in more complex(i.e. non-static page) situations?
Should it just be the intermediary between the users and your database process? Heck, should the database process also handle all the formatting of the data as well as the queries? That is fairly CPU intensive(and doesn't scale well at all), especially if you're getting to larger data sets.
Point of this whole thing is that you cannot reasonably do complex computations in Node. While most "Web 2.0" applications probably does not need to do that in response to request, there are also many web applications that actually do something that might be reasonably called complex cpu-intensive processing in response to user request (although most such things are hard to scale). So it really boils down to "right tool for the job" - and for many practical (but maybe un-popular) problems, Node simply is not the right tool.