In Europe, fintech has a lot of legacy systems in COBOL, with no plans of getting rid of them anytime soon.
Also a lot of languages are starting to adopt features from the functional paradigm and it's always good to know where they came from. It's unfortunate that many engineers use techniques like memoization (via annotations for example) without understanding their underlying principles.
As for significant systems, Emacs is probably the most well known. Also the Clojure community is very active.
Op is posting about learning cobol for job prospects, so an esoteric fad language without job prospects is probably not the language they are after.
Also Lisp is far from an "esoteric fad language", it's been around since the 1960s, has both ISO and ANSI standard dialects, and has some significant usage in industry. Like COBOL most of the companies that use Lisp are using it in specialized situations (I would argue this is similar for most functional languages). I feel like it's not talked about as much as COBOL because for COBOL specifically there is a market demand to maintain legacy systems that outstrips the supply. Because of this the role of maintaining these systems - at least historically - paid very well.
This is a pretty gross oversimplification and it's also a mischaracterization to say it's "a more widely used and active version". I wouldn't even put it in the same catagory as traditional lisps since those will compile down to executables and clojure run in the JVM
They "kind of" do. `:executable t` does not actually make an executable in the way most people understand it. It's a big blob with lots of stuff you would expect to be stripped out in other languages. Short of paying several grand for Lispworks, you are not going to make true executables in something like SBCL.
In contrast, Clojure with GraalVM is fairly easy to do and will give you smaller binaries.