Right, but it's the 'every once in a while" part that I wonder about, i.e. the variance of that distribution. Maybe you could argue that since the Maxwell-Boltzmann distribution is narrower the lower its mean, if kT is small then it's an exponentially tiny effect.