At scale do you think this would cause cost issues with increased IO across datapoints when pushing refreshed data?