it might be good to try to come with a more systematic way to handle the crashes resulting from this