I'd rather a Bitcoin that is both absurdly valuable, and cheap to use, because of high transaction volumes and adoption rates.
Cheap to use shouldn't mean mean free or almost free.
Bitcoin is now past the era of needing to give out free samples and rely on word-of-mouth marketing.
All the right people already know Bitcoin exists and what it can do.
At this moment, we need to stop subsidizing transactions and see what price the market will bear in a 1MB environment.
Innovations and substitutions will ensue, and we may find we requires less (or more) than 20mb.
Let's give it some time and observation before fixing what is already working spectacularly well.
Cheap to use or almost free, only hurts miners if there is an artificial cap on the number of transactions.
If instead the number of transactions is based on supply and demand such that no artificial cap affects potential revenue then miners stand to make more money. The consumer does not care that the $3.50 coffee cost you, say $3.5001. The billion or so transactions that paid $0.0001 to be included in the next block mean the miner that solved it gets paid $100,000.
So it can be both 'almost free' for the end user, and incredibly lucrative for the miners. All at the same time.
I agree with your premise to let the market decide. If the protocol stops them selling goods (space in the next block) that's not really a 'free' market though is it.
(*all values converted to representative dollar values for illustration purposes - in a mass adoption scenario, everything would be in BTC anyway)
Is it worth self-exile from TOR and other low-bandwith networks just so Bitcoin can attempt to track every coffee bean/bus ticket and be a single point of failure for all transactions great and small?
The market usually prefers different mediums of exchange for transactions of very different size, and associated costs/fees/friction increase according to value transferred: