It helps to think of it as a tragedy of the commons: absent of a blocksize it will eventually become in the miner's best interest to include as many transactions as possible in their blocks and consequently restricting access to governance of the network by way of bloating the blockchain.
'Yes' on the tragedy of the commons comparison at first, but 'no' if you think it through a bit more...
Only the biggest pools matter anyway, both in a vote and a self-enforced scenario I describe. Let's call them A, B, C. A vote to hard enforce scarcity will only succeed if A, B, C individually want scarcity anyway, so I'm allowed to make that assumption for my scenario as well.
Now, your objection (ToC) is valid in principle, that it is indeed in each of A, B, C's personal interest to include as much as possible contrary to the larger scarcity goal, but I point out that in this particular scenario, self governing is less likely to fail because (a) cooperation needs to be established only among a small group of actors (i.e. pool operators) that directly control the majority behavior, and are able to directly communicate with each other (as opposed to a large, amorphous population that usually underlies ToC type scenarios), and (b) because defection is publicly visible.
If, say, the two or three largest pools really
want scarcity, they could try to cooperate, set a limit, and self-enforce it. (yes, free market lovers, that's a form of collusion). Blocksize is public, so cheating is possible but visible. If cooperation of the relevant majority fails (which I consider unlikely, as per the argument above), it makes sense to talk about voting processes and hard coded limits, but not before it has been established that this type of cooperation cannot be handled by individual actors.*
* Yes, that last one is an axiom of sorts, not an empirical claim. Something like, "don't solve with laws what possibly, in most cases, can be solved by individual rationality". I'd suspect most 'coiners are not entirely alien to the idea though.
(EDIT) Perhaps I should add that I am not going against
some hard max value in principle. We have one now, we should keep one (only higher). What I'm trying to argue for is that miners only need this value for its originally intended purpose, to enforce a reasonable upper limit for
technical reasons, i.e. as a measure against spam attacks. I reject however the notion that the
economical aspect of that value needs to be included in the current debate, as any actual value below below the technical max has a good chance to emerge as a cooperation effort if there were an interest by the big miners to do so, as I argue above.