A max block size is needed because of the asymmetric incentive structure. -- There is no increase in revenue for node running with larger blocks, only increased costs.  Under free market conditions, the larger the block chain data set gets, the worse our node/miner ratio should be expected to gets, and we should expect a declining # of nodes overall, even with increased mining.
This is a bad result for network resilience and performance.

But... that is just the benign bad effect, one of the pernicious risks that would aggregate (with an indefinite and inaccurate max block size) would be an increasingly easy attack on this pain point.  Eventually, someone who wishes bad things for Bitcoin could mine fully padded blocks of transactions to from and to themselves and increase the costs for everyone else at no marginal cost to the miner.

If the economic incentive structure were fully resolved (and I don't think that we know how to do that yet), there might be no need for a max block size!

The root of your question is another problem, I've discussed elsewhere.
For so long as we are stuck with some hard-fork based solution, we ought also set forward a criteria for changing it.  Currently it is structured in a central banking type small group of deciders which is going to be a source of criticism.
What is needed also is some mathematical method of determining when it should be raised (or lowered).  Getting this right would give us the "accurate" part of the solution of not being both inaccurate and indefinite.

So... assuming there is a problem, the current proposal is not the solution.