Thanks for the investigation, Richard Walker! As you say I'd prefer to leave things as they are then, so to not change semantics of the existing constraint.