That's true. But money in that micropayment sense might not be entirely insane (assuming the whole micropayment model isn't entirely insane to begin with). The idea (as I understand it) is that you'd get a tiny wee payment every time someone quoted your work; so why not every time someone quoted your code? In the context of some Xanadu-like system, wouldn't that make sense?
But (I think) he means as a product, not a service (i.e. for bounties, only the asker pays, so it's a service; though it's true that everyone else benefits too).
I like this idea. The big problem, as always, is the boundary or interface of the module/component. (1). The buyer has to specify it, which is usually the difficult bit and writing it is usually easy. (2). For a market, you need to find modules that many people need, that people can search for and find, and that are difficult enough to implement that people would buy rather than build.
Maybe it's possible, for things that seem one-off or too specialized to a person, if you reduced friction enough, in a SO way. They've solved the search problem.
I guess an example would be a square-root function. I wouldn't know how to code it, I'd have to look it up, test and debug. It would take at least hours, I'd think. Of course, it's such a useful one, it's already done. That's the fate of all generally useful functions. The exceptions are new ones and niche ones.
I guess you could just add a little "debugged code" box for SO, and the person keeps coming back to it, to support it. This might be a useful addition; but I think you're right that money would spoil it. However... if you imagine a spectrum from one-liner to massive library, there's probably some level where the amount of work would make it reasonable to pay.
Conclusion: to see the value in the idea, leave money out (build something people want). How would you modify SO, or start a new one, to provide this value, or reusable functions/code snippets?
Matt Cutts (Google SEO) fondly referred to this site as "Expert-SexChange" at Wordpress Conference in San Francisco in 2009! Google it. The video is funny!
But it makes me wonder when exactly we start to call something over-engineered. The same function will be implemented differently depending what you are going to feed into it i.e. if the numbers are sufficiently large then you need to worry about integer overflow etc. Probably a function definition should be considered incomplete without pre/post conditions.