augmented image datastore can't handle multiple observations - Is this a bug?

My goal is to generate an augmentedImageDatastore with

- inputs of type 160x160x5 double (5 channels of images with 160x160 pixels each) and

- responses of type 160x160 categorical (pixel-wise labels)

The image Matrices are given directly, not generated from images in a folder. Unfortunately using the augmentedImageDatastore with more than one observation fails:

Self-contained example code:

%Following four lines only for testing, data is already given that way
input = zeros(160,160,5);
response = categorical(zeros(160,160));
input = {input;input};
response = {response;response};

%Generate table consisting of input and responses
tbl_tworows = table(input,response); %table containing two samples
tbl_onerow = tbl_tworows(1,:); %table containing one sample

%augmentedImageDatastore with one sample works
aug_onerow = augmentedImageDatastore([160,160,5],tbl_onerow);
aug_onerow.preview() %doesn't fail

%augmentedImageDatasore with two samples fails
aug_tworows = augmentedImageDatastore([160,160,5],tbl_tworows);
aug_tworows.preview() %fails: "All input variables must have the same number of rows."