If (the amount of unconfined transactions > some limit && the last 2016 blocks contain 9x% of 1MB worth of data)
decrease the difficulty by x %
This means:
Faster transactions if needed
More transactions if needed
More inflation if needed
The block reward driving mining costs will be obsolete sooner and the 21 mil coin limit can remain *only it will happen sooner
This method provides an economic incentive for large miners to spam the blockchain with many very small transactions, which is one reason why it will never work well.
Interesting, I haven't thought about that. But I think this incentive can be removed if the difficulty reduction is tied to the transaction fees somehow, like only counting unconfirmed transaction with some set amount of fees. This would still push out zero fee transactions, but well to be honest those are only there because of the block reward in the first place and would become unsustainable sooner or later anyway.