The problem may be with the fuzzy conceptual notion of "systems language" that is the starting point of many of these languages that end up with effectively conflicting set of goals.
(I agree with Animats comment elsewhere in this thread blessing all threaing models as "OK in isolation". It's just they don't play well together in a single language.)
The OSes designed at Xerox PARC, ETHZ, Microsoft Research, Olivetti, UK Navy, some of which used for quite some time in production, were quite clear, there wasn't any fuzziness on them.
(I agree with Animats comment elsewhere in this thread blessing all threaing models as "OK in isolation". It's just they don't play well together in a single language.)