Cost of prying a favoured text editor out of a developer's fingers? Possibly more that $4. If Jon could sell 135,000 licences a year at the current price, he'd make the same. Also, the product is worth the money, and $4 is a terrible signal to the market. It says 'This product is shitty and valueless. Please look elsewhere'
One interesting organising principle, from open-source text editors, is emacs. The core code of emacs was controlled with an iron fist by Richard Stallman. It was massive work to get something into the emacs core. But developing your own major mode (say, HTML editing) was something you could do yourself, and distribute over the web. The reason you stay with emacs, if you're a fan, is because it has good modes for the particular language you're developing. Come for the text editor. Stay for the packages.
So emacs has a tightly controlled core, with a thousand-flowers philosophy around plugins. That seems to me a strong strategy, and it relies on one central developer, making sure the product keeps its integrity while many people experiment with new technologies. It's the world we have with Sublime Text today.
So I really do think that Jon should not open-source the project, unless he decided he's really done. Rather, he should keep control of the core, develop core packages for things like python and C and HTML, and then let other people integrate with new tools and such. Pick a new tool -- say, the Mercurial distributed source control system. How many text editors out there have integration with Mercurial? Few, I imagine. Sublime's core should offer all the API hooks we need to write our own Mercurial package, and that should be open-source.