test_axes.py contains a few tests of the form

def test_fill_between_2d_x_input():
    x = np.zeros((2, 2))
    y1 = 3
    y2 = 3

    fig = plt.figure()
    ax = fig.add_subplot(211)
    with pytest.raises(ValueError):
        ax.plot(x, y1, x, y2, color='black')
        ax.fill_between(x, y1, y2)

and similarly for fill_betweenx. The ValueError actually gets raised by plot, not by fillbetween, which is almost certainly not the intent. In any case, this should be easily fixable by getting rid of the call to plot. The tests could probably be parametrized as well.