I'm not surprised that it also applies to code. However, so far it seems to be language dependent: Also interesting is that spoofing is possible. And that implies that code stylometry can be readily defeated. Also, regarding prose in different languages, they say: OK, so I'm somewhat active online as my meatspace identity. But in my native tongue, and not in English. But I certainly don't just translate stuff. It's almost like I'm a different person when I'm thinking in American English. I draw on usage and slang that I've picked up from friends and coworkers, from many places. From "all y'all" (Southern US) to "gobsmacked" (British) to "take the decision" (Mexican). And that's totally absent from prose in my native tongue, so I doubt that stylometry would link it to Mirimir.
I suspect this would be programming language and IDE dependent. For the coding I do, the stylistic aspects are in fact constrained by the IDE, architectural, component, library and style guidelines imposed by the projects - so relatively little to go on there, particularly when others are doing the same thing. I'd view what I do as more being a librarian and hooker-upper rather than a coder. Of course, stylometry on natural free text is far more likely to be more individual, for example, in the use of pronouns. As usual, the article is somewhat coy in quantifying false-positives, and obviously its reasonable success rate is only when predicated on a known pool where previous example of work is already known.
They do note that those who code based on searching and combining stuff are harder to identify. As opposed to those who write from scratch.
