Oh, I'm not arguing that its not in use - I'm arguing from the point of view of the users of the software itself (not its authors) who may want to write extensions. I feel users are unlikely to jump through hoops to write extensions to software, yet I feel that Guile creates hoops for a number of reasons, including (but certainly not limited to) an aversion to lisp (i have personally observed this), unfamiliarity compared to the users every day language, the difficulty of a functional-focused vs imperative focused language for quickly hacking an extension together and ease of obtaining documentation.
I think we agree then. The (to most people) unfamiliar syntax and programming paradigm are certainly barriers to entry. It's not what people expect and I doubt most people writing plugins for programs want to do so in unfamiliar territory.
[1] http://lilypond.org/doc/v2.14/Documentation/extending/scheme...
[2] http://www.geda.seul.org/wiki/geda:hse_howto