I'm OK with a smaller size + a formula that ramps it up over time. We are far from having enough demand to fill 10MB blocks, let alone 20MB today.

To put it in perspective, to be feeling squeezed inside 10MB within two years, we would need to double usage five times. I wish I knew a way to make that happen. So the chances of us going to 20MB blocks full of real transactions any time soon is close to zero short of some amazing killer app that takes the world by storm (in which case: yay, nice problem to have). As long as capacity significantly outpaces organic growth, we should avoid problems.

The reason to pick 20MB then is merely one of expedience: we have to pick a number, 20 is tested and seems to work, and we don't want to get caught by surprise if demand does outstrip expectations.

Still, I question the underlying logic. We have no idea what connectivity into China will look like a few years from now: it's seems to be a function of politics rather than hardware trends. It might go down rather than up. So 10 vs 20 feels a bit arbitrary. We can't let the Chinese government dictate how Bitcoin is used, that would never be accepted by the rest of the world. But if we optimistically assume things don't get worse, and 10 == more acceptance, then alright - it should make no difference in practice.