A list containing 5 elements has a 1/(5!) (1/120) chance of being shuffled to the same order. This was causing unit tests to fail locally by pure fluke. Tests now generate a list from 1-52, which has a 1/(52!) chance of shuffling to the same order, so the chance is basically 0. If these tests ran once per second, it would take 1.9x10^50 * age of the universe to hit a collision.
This still can happen. Randomness does not guarantee we won't hit the same order. The odds are just so astronomically small that it's not worth considering. And if the CI run fails due to these tests failing, just re-run them anyway and call it a day
