I think an important difference is that when you shuffle them, the unfairness stops being correlated across multiple assignments, so the "aggregate" unfairness over the course of the semester is much lower.