The proposed approach is allarming in a way, but if you see it in a unit test way, it is ok as long as there are sanctions on the provider when a client realize (using input and output) that the software is not running as advertised