Consider a class T that implements the __array__ protocol with shape (0, 0) and has length 0, so it is consistent with itself. When calling np.array([T()]), the result has shape (1, 0); 1 from the list and 0 since that is what T-as-sequence reports as its length. But should the result shape be (1, 0, 0), i.e. should the __array__ protocol take precedence over the sequence?

This came up in #13659 when discussing the case of pandas.DataFrame() which exhibits this behavior. Below is a class taken from the test to fix that issue.

        class T(object):
            def __array__(self):
                return np.ndarray(shape=(0,0))

            # Make sure __array__ is used instead of Sequence methods.
            def __iter__(self):
                return iter([])

            def __getitem__(self, idx):
                raise AssertionError("__getitem__ was called")

            def __len__(self):
                return 0