According to what I have read, there are models that, when faced with difficulties and unable to find an answer, simply stop trying. This makes sense, because it requires much more processing (perhaps in vain, for the moment).