Among the ML descendants, I suggest Scala 3, because it has essentially all the power that we like in Haskell, (HTKs, good support for ad-hoc polymorphism), but runs on the JVM, a mainstream platform with a vast library ecosystem.
Your point about the vast library ecosystem might be valid. Personally, there has only been one time in several years working with Haskell I’ve wanted to use a library that I couldn’t find an analog to in Haskell. It was WeasyPrint which is a Python thing, and there was no problem to run it from my Haskell program as an external process.
Although thinking about this some more, I probably could have just used Pandoc…