My take on the bitcoin unlimited idea:
It is best to have just an unlimited blocksize, and a parameter limiting the max block size to produce, information about what to do, shared outside of the blockchain.
In this case, the miners always are the ones to choose the actual block size. They will choose according to their best interest, which is basically the cost and the advantage, but remember that in the market, it is the long time psychical advantage and cost we talk about, which includes considerations of what is best for the system in the long run, or if you take it all the way, humanity. It is not required to take into account other things but your self interest, but it is allowed (and if you ask me, it is not an insignificant part of the free market).
So to spell out the most important factors, the advantage of making a large block is obviously the collection of fees, but also the future value of the miner's own coins an his business, because large capacity is needed to make the system to be sound money for as many people as possible.
The cost is the relatively small but not zero added cost of a large block in hardware, power, network and administration (thinking about the size), but maybe the largest cost is the risk of the block getting orphaned. For a business, risk transforms directly into cost that goes into the bottom line, the business problem is to manage that risk.
So an "unlimited" miner will have to consider which is the largest block he can make without risking too much on his bottom line. Voting with version numbers and comments in the blockchain can be of some help, but he can never fully trust it. In the end he must take all available information into account, which includes what is talked about in the social arena.
To explain this with an example, if a limit of 8MB is the current relative consensus, diverse sizes appears, in general increasing, and when the first 6MB block appears, there will be a need for a new consensus. Things that can be of importance, are:
Have some nodes crashed at 6MB? Have some nodes given up? Have we had miner centralization of importance? Have users been hurt? Have some wallet software or servers met problems. Have businesses had problems. Do we have lots of microtransactions that are not worth carrying? If there is not an apparent consensus, the risk of orphaning is bigger, and the miners will have to take baby steps going forward. I am sure hundreds of other more or less relevant questions will come up, so don't take this example too serious.
The important part is unlimited size on verification, miners decide block size at creation, and out of band information exchange.