On the complexity of second-best abductive explanations