Yeah it's a pretty tough question! I've resisted doing that until I had a relatively high certainty that their published results contained minor mistakes, which I assumed they would want to know about. I've also been explicitly apologetic and tried to keep it super brief.