It's not something I've explored or really care to pursue, but I've sat though at least a half dozen talks on audio synthesis in Clojure including one by Rich himself. They do signal generation in Clojure that feeds data into another system that does the real-time portion.
I agree that Clojure isn't suited to hard real time systems and it's not hard to come up with drawbacks to the above approach but you just happened to pick an example that a lot of people in the Clojure community seem to care about.
> feeds data into another system that does the real-time portion
Well of course that's a technique but "another system" isn't Clojure. It does so happen that a Lisp language is especially fluent for programming audio processors because of its compiler-like expressions and macros.
I agree that Clojure isn't suited to hard real time systems and it's not hard to come up with drawbacks to the above approach but you just happened to pick an example that a lot of people in the Clojure community seem to care about.