It is an interesting question, because how does one generate velocities at > 0 K? > Random sampling from a Maxwellian distribution with mean zero and zero width :) At the end of the first step all velocities will be scaled to zero. That shouldn't be very unstable. Cheers, Tsjerk