If Mike Pall is to be believed, dynamic languages do better with specialized environments, which is why he thought LuaJIT 2 was a better strategy than Lua on the JVM or LLVM. SBCL may or may not incorporate similar optimizations, but it's definitely built to handle Lispy weirdness from the ground up.

There was briefly an attempt by Raphael Amiard to compile Clojure to Lua, which could also test this theory, but it was never completed and I doubt anyone else feels like doing that much work for free:

https://github.com/raph-amiard/clojurescript-lua

> compile clojure to lua

That's fennel: https://github.com/bakpakin/Fennel

Writing your own language is more characteristic of the kind of work people do for free -- unfortunately, it creates a high barrier to entry. You can't tap in to Clojure's existing library ecosystem, for example. Does Fennel have instaparse? Leiningen? (And sure, you could use Lua libraries, but they're built to work like Lua libraries, so at that point you might as well be writing Lua.)

https://www.jwz.org/doc/cadt.html

Of course this same critique could have been leveled at Clojure from its outset, but Hickey has really stuck with it for the long haul and made Clojure a very good language in itself as well as just being a Lisp on the JVM (and it even has sort of an ecosystem, a rarity among Lisps and what made me interested). I thought a few times about writing my own language when I was younger, but I realized it's a life-consuming task, and not something you can just do. But I do wish the author the best.