I've moved to Scala in 2012 (because I needed to work on the JVM) from over a decade of Ocaml. I found that the absence of full type inference was not a major problem. The JVM library ecosystem on the other hand ... lack of libraries (at the time) really hampered Ocaml.

Does anyone use Scala in cases where OCaml is a viable option? My impression has always been that Scala is "better than OCaml if you need the JVM"

Mainly Spark, but it really depends on the domain, what libraries exactly that you'll need, hell I look at what kind of deployment routine the client wants, what their maintenance crew looks like compared to what they want from me.

There is plenty of scala in fintech, I can tell you that. But there's plenty of ocaml, f#, matlab, excel macros, r, kdb, cobol, fortran, c++, a vast array of different technologies.

Most places that choose scala are choosing it over erlang or Java or c#, they USUALLY only choose scala over ocaml if they want to have some old ocaml code talk with new infrastructure that they want built in scala, ime. Because of how performance works and that multi core isn't working yet, ocaml is limited in application if they're willing to find or already have scala programmers.

I don't like having to give up full type inference, but it's much better than having to give up HKT.

You can get close enough to HKT with OCaml that you don't really feel any pain there. We have our share of generic monad libraries.

Up to a point. I had a look and an implementation of traverse is already getting pretty cumbersome, and something like recursion-schemes seems beyond what anyone's even attempted.