Hmm I have come across many a codebase in countless languages where there was no way to discern that type of information from the naming convention. Even accessors in say C# or Java are perfectly legit to do any side effect they want. In fact I think it is worse in those situations - because side effects are not modeled. Indeed, no language has found a way to model non-termination in the type system, hence the halting problem.

Honestly I just haven't experienced this very often in Haskell... When I do it is usually a 'doh moment and a simple fix.