If you have rough estimates for what the upper limits of this approach are (based on your tests), I'd love to get a datapoint to keep in mind. I use this at low-moderate volumes.