I thought using random.seed() would fix the problem by making the tests repeatable, but in #626 changing a linein README.md caused the output (on travis) to change.
Am I using the random.seed incorrectly or are there other factors at play?