We ran each method in under 24 hours on a singular H100. I understand your point and think we will include this in future iterations of our work since this is very interesting from the user perspective. Though, in the paper we focus more on algorithmic concerns.