Re the experimental version of MLton for Darwin, it might be worth E-mailing Reppy, or McQueen at U of C. I remember that they seemed quite interested at some point.