I would suspect that there's some sort of internal algorithm used to select the best sort algorithm for a given data set, based on the data type, data volumes etc