I've always been curious about Jane Street. It seems like a too-good-to-be-true story. Picking an obscure but powerful alternative to C++ and Java for a highly-competent core team makes sense, and has been done by several banks. But hacking OCaml does not fit the personality profile of (m)any traders I've met — most will happily whip up a spreadsheet to help their work, but writing extensive software and learning Hindley-Milner type systems falls outside their ordinary needs or interests.
Well, HM is about type inference and they can just assume it works correctly. And I don't think they are expected to write extensive, complex programs in the language but more put together pieces from their "vast trove of proprietary libraries", which are most likely designed just for the fact of faciliating the use of the language for users who are not (primarily) software developers.
I'm not sure you can realistically wire together pieces of functionality written in a statically typed language without understanding its underlying type system. Sure, by the time it compiles without errors the code probably needs less debugging than in a dynamically-typed language, but getting to that point can be difficult and frustrating. (Speaking from experience as a TA.)
You don't have to write extensive software or have a deep interest in type systems to use ML languages. You just have to know enough of the syntax to understand the API and write some functions to work with provided DSLs.
