I don't think this really improves the situation? With this PR

scatter([1], [2], c=np.array([[120,130,140]]))

gives

ValueError: 'c' argument must be a color, a sequence of colors, or a sequence of numbers, not [[120. 130. 140.]]

but this value of c is a sequence of number in the sense of scatter (which will flatten them): for example

scatter(np.array([1, 2, 3]), np.array([3, 4, 5]), c=np.array([[120,130,140]]))

works just fine.

Edit: edited incorrect example.

If you look at the docstring, in particular the note section, you will find that in the example in question, and in the test I added, c is not supposed to be flattened, and indeed it isn't. Therefore the change I made brings the error message into sync with the docstring and its intent, which is that if c is a 2-d array, and it's first dimension matches x.size, then it will be treated as a sequence of RGB or RGBA--or at least the attempt will be made. At that point, color mapping is out of the question, and the failure is not a color mapping size mismatch.

Right now the docstring reads

         c : color, sequence, or sequence of colors, optional
            The marker color. Possible values:

            - A single color format string.
            - A sequence of colors of length n.
            - A scalar or sequence of n numbers to be mapped to colors using
              *cmap* and *norm*.
            - A 2-D array in which the rows are RGB or RGBA.

but that's not really true: something like

scatter(range(6), range(6), c=np.array([[10, 20, 30], [40, 50, 60]]))

and there c is 2D and gets flattened.

Thus, if we take the docstring at its word, for the purposes of scatter, "sequence of n numbers" (to be mapped to colors) includes 2d-arrays that get flattened. I don't agree with this choice, but that's irrelevant here.

But then the error message "'c' argument must be a color, a sequence of colors, or a sequence of numbers, not [[120. 130. 140.]]" doesn't make sense, because [[120. 130. 140.]] is such a "sequence".

Or you need to fix third item of the docstring (perhaps by copying/lifting part of the notes section there).

This isn't worth any more time.