Well, this isn't a very technical discussion, but I can quickly list them: for shared mutation I think the Clojure approach of transactional/atomic mutation everywhere rather than RT-everywhere is simpler; for correctness/side-effects, imperative effect systems may be easier (well, they don't quite exist other than in a very limited, experimental form, in Koka and Java 8's pluggable type systems -- but Haskell doesn't provide a very fine-grained side-effect-control either). For correctness, Esterel (which has been well proven) and its descendants, seems a lot more sophisticated in terms of proving correctness, while being more intuitive (of course, Esterel introduces a limited computation model, but so does Haskell -- or languages in the same vain -- if they are to be decidable).
> the problem for sharing that experience is many of them are in internal proprietary systems
Yes, that, too, is a problem, but the bigger problem is that uses in the industry are vanishingly rare -- and the bigger the domain challenge, the rarer still they are. For example, I don't think there exists a single case of a Haskell program that tackles complexity in such a way that implementing it in, say, Java or Clojure, is prohibitively expensive.