Being able to use Clojure would be great, but as you said not a viable option yet. GC pressure is the biggest problem, and it's one that sadly won't get solved too soon.
I've also looked at kawa (scheme), I might try that.
I've also looked at kawa (scheme), I might try that.
My understanding is that there is greater interest in making Clojure more suitable as a library and with a leaner runtime requirement. If you have a situation were you don't need a dynamic environment, having the option to compile functions to static methods instead of classes should result in a big performance increase.
I have only briefly looked at Kawa Scheme, but I have not had the time to do much with it beyond a "Hello, World!" sort of project. Mirah is another interesting option. It uses a Ruby-inspired syntax but requires no runtime library.
I certainly welcome any improvements in Clojure.
I've looked at Mirah, but it's quite immature and has serious problems with Java generics because of Java's type erasure.