Biggest problem I had with Akka was that it was that it worked better with Scala than Java. As Java recently has re-taken most of Scala's momentum its a bit awkward fit.
Same story with .NET - less and less reasons for F# to exist with each new version of C#. Only F# doesn't even bring any new useful frameworks to the table.
Sad to hear that, F# is an excellent programming language. Scala has matured a lot, there is no "better-Java" anymore, that's the reality, there are three ecosystems today, Akka, Typelevel and ZIO. The competition between these is also seen in the social media and other mediums. There's also Clojure, which in my opinion is still an uncharted territory.
Using anything build in Scala from Java is abysmal. Have fun with no default parameters, creating Seq$.MODULE$ everywhere etc. I'm not from US, so I've never have seen so much dollars anywhere.
I had a little success by simplifying the interop layer. If you can wrap your scala code in simple classes that expose only Java-compatible structure, then you can remove additional complexity of java into scala interop.
You might be able to write a shim layer in Scala. Java interfaces, Scala classes that implement those interfaces and call the third-party code. That's the nicest way to do it IME.
I think that's what we plan to do. Still, it introduces some additional problems: we were able to target all the supported Scala versions with one package, now we'll have to do separate ones for 2.11, 2.12, and 2.13.