Barring that, "Successor" ML is cleaning up some deficiencies in the original standard, and ML implementations are starting to integrate the changes.
But yeah, a strict ML with the pristine syntax of SML; the modular implicits coming to OCaml; the innovative ideas in 1ML -- now that would be a great language.
Sigh, back to Scala...
As someone who did a lot to market certain OCaml libraries, I have seen similar interest in some of the "modern" SML libraries I have played around with. Not a ton of work has gone into both of these, but you may be interested to look at two projects MonoML[0] and SSMLS[1].
These are some /very/ minimal projects I've worked on in SML in the past few months to gauge how well SML might work as a modern, web-friendly, higher-level language. The results are certainly not completely disappointing.
Edit: Oh also! HamletML[2] (yes, written by the 1ML guy) is a very interesting ML implementation that would probably be a great way for work on modern language features. Not that I know what I'm talking about, but it could also provide the framework for an even more powerful backend/optimizing compiler.
[0] https://github.com/eatonphil/monoml
Adam Chlipala has a writeup[3] on comparing SML and OCaml that may be of interest to the discussion.
[0]: https://github.com/xapi-project
[1]: http://anil.recoil.org/papers/2010-icfp-xen.pdf