It is not particularly likely to happen early either because ... the diff adjustment never gets us down below 6/hr for any length of time
So doesn't that mean that reaching 21,000,000
is likely to happen early, because the blocks per hour is averaging 9 or whatever?
The rate of generation is somewhat arbitrary anyway, and whether 21 million is reached a decade or a century after I'm dead doesn't make much difference to me.
Within a couple of decades we'll be past 20 million. After that time my guess is that the rate of coin loss will be a bigger factor than the rate of coin generation.
Just speculating here.
When difficult "overshoots" we will pay back a bit more time than the comparable sized "undershoots" that we've been doing have gotten us ahead.
If we do 9/hr (50% over) for 2016, then 4/hr (50% under) for 2016 this takes 224 + 504 = 728 hours which is longer than 336 + 336 = 672 hours.
So essentially were just in an extended period where we haven't paid any time back and it might not come for a long time.
I'm not really sure, and I don't have a solid conclusion here. I think smooth growth in power gets us ahead, but jerks up and down get us behind. Maybe someone can write a simulation (or just do better math than me).
