My suggestion for the Raspberry Pi Foundation is make like 10x the amount you THINK you need up front.
Nice idea, for us. Let's think this through...
I see a number of problems with your idea:
1) The Pi factories can only make so many Pi per day. If all that production were switched to a new model prior to launch then there would be no production of existing models during that time. That means nothing to sell in that time. That means no income. That is bad.
2) Your suggestion requires increasing the delay between finalizing a new model design and launching it for sale. Time to make that "10x" amount. We would have had to wait longer for the Pi 4 launch. That is bad.
3) Making 10x of stuff just to hoard until launch means holding on to stock until launch for longer. That increases storage and other costs. That's bad.
4) There is the risk that having made 10x, more than expected demand the device is flop in the market, nobody buys it, you have wasted a lot of money and time on building the. That's bad. You have lost income due to 1) above. That's bad.
All in all a bad idea to over stock prior to launch.
Memory in C++ is a leaky abstraction .