Yes, the "best in class" libraries change every few years in Clojure's universe, but they always bring huge improvements that require completely new APIs.
For instance:
https://github.com/plumatic/schema -> https://github.com/metosin/malli
https://github.com/juxt/bidi -> https://github.com/metosin/reitit
https://github.com/bhauman/lein-figwheel -> https://github.com/thheller/shadow-cljs
https://github.com/stuartsierra/component -> https://github.com/weavejester/integrant
https://github.com/technomancy/leiningen -> https://github.com/practicalli/clojure-deps-edn
> So overall Clojure is in a strange situation. For picking it up on-the-job its not the responsible choice probably for most situations I work in. For my private tinkering the JVM dependency feels just a little too heavyweight. I don't see that Clojure will hit an inflection point of adoption in any way in the future that would make it a viable candidate for me to use at work.
I think about programming languages as tools for different bottlenecks.
– Computational: Rust
– Networking: Go
– Business logic: Clojure / Python
And for many people Python has everything that Clojure is missing: a very low barrier to enter, and a very stable and well-documented ecosystem of frameworks and libraries.