Wolfgang,
Haskell is non-strict but not necessarily lazy. So it's possible that an expression is reduced to WHNF although it is not used yet. Could this early reduction also happen to outputs of unsafeInterleaveIO actions (which might trigger the action too early)? While I'd expect those outputs to be evaluated lazily (reduced as late as possible), I cannot find anything in the docs that guarantees this.
"unsafeInterleaveIO allows IO computation to be deferred lazily. When passed a value of type IO a, the IO will only be performed when the value of the a is demanded. This is used to implement lazy file reading, see hGetContents." http://haskell.org/ghc/docs/latest/html/libraries/base/System-IO-Unsafe.html... Is this the kind of guarantee you're looking for? I'd bet against getting any decent durable, portable guarantees for any unsafe* functions in Haskell; but the above behavioral description may be strong enough to suit you.
In addition, I'd like to know whether unsafeInterleaveIO outputs are guaranteed to be evaluated at most once so that the interleaved action is executed at most once. Again, I suppose that this is the case while I cannot find a guarantee for it.
I'd be surprised if an implementation didn't have that behavior. I'd also be wary of anyone claiming to guarantee it, beyond compiler X version Y. John Dorsey