When the whole feature lands, we can test that the UI shows at the right time, and behaves as expected.  Perhaps a pixel test would make sense, but maybe testing updates to the data model separately would be less of a "detect changes" test.