We have a (temporarily) failing test that checks that if we delete a
corpus, the associated context (i.e. document) is not deleted with it.