It wasn't immediately clear that the proper random seeding was taking place. This ensures that the same plaintext encrypted twice does not result in the same ciphertext. It will also be a good test to keep around incase of future regressions.