I'm a big fan of adding noise to systems to make them more predictable (it was the topic of my thesis: http://tlb.org/#papers http://tlb.org/#papers). But, I be

I'm a big fan of adding noise to systems to make them more predictable (it was the topic of my thesis: http://tlb.org/#papers http://tlb.org/#papers). But, I believe it's not needed here because the arrival times of /newest voters are already sufficiently noisy. You only need random noise of +/- 0.5 to get predictable results, and voter arrival patters give more than that already.