Some obstacles in characterising the boundedness of bi-parameter singular integrals