Good point. I'm very curious now about the selector overhead. We do a number of operations across varying intervals. Surely one of the goals behind distributing

Good point. I'm very curious now about the selector overhead. We do a number of operations across varying intervals. Surely one of the goals behind distributing the timestamp updates is to minimize the time we block this pipeline. That being said, it would be interesting to benchmark a variety of update interval and batch size combinations.