It's just an infrastructure problem. Hardware scaling is hard to get right. So it's easier for the engineers to have it released by small increments. This way you can slowly adjust hardware as the need arises. If any problems show up, it will be easier to track it down and fix it.