While I doubt it's an issue in practice, I have a real hard time believing that a compromised source of randomness actually improves the function. If it does, t

While I doubt it's an issue in practice, I have a real hard time believing that a compromised source of randomness actually improves the function. If it does, the algorithm is presumably not well designed in the first place - which would be similarly concerning.