Lambda the Ultimate

I'm under the (possibly simply clueless) impression that skeleton libraries for parallelism e.g. OCamlP3l are a good way to be highly safe and avoid weird parallelism bugs in your system - if I understand it, the point being that the skeleton library has pre-tested the skeletons to make sure they work (modulo the user really doing something clueless with them). But I don't get the impression this approach has taken off. Is this due to the inflexibility which recent OCamlP3l's claim to have at least somewhat addressed (perhaps also SuperPAS)? Or is there some more basic problem with the whole idea? Is it wrong, or outdated, or ...? I'd very much like to hear anybody's experiences with such things. Thank you.

(P.S. and how they feel compared to e.g. the Actors approach, or the GdH approach, or others.)

(P.P.S. and how different takes on skeletons feel comparatively e.g. Java Lithum, DataRush, or any other such thing.)