I think the first two examples relates to Deep Learning models. With regards shuffling data, it depends on what your looking at. For example, for time series based forecasting, the data should not be shuffled.