The idea is to avoid a "tsunami" of error messages when a heavily used theorem breaks in the beginning of the file