I worry a lot that there is a widespread misunderstanding that blocks being "full" is bad-- block access is a priority queue based on feerate-- and at a feerate of ~0 there effectively infinite demand (for highly replicated perpetual storage). I believe that (absent radical new tech that we don't have yet) the system cannot survive as a usefully decentralized system if the response to "full" is to continually increase capacity (such as system would have almost no nodes, and also potentially have no way to pay for security). One of the biggest problems with hardfork proposals was that they directly fed this path-to-failure, and I worry that the segwit capacity increase may contribute to that too... e.g. that we'll temporarily not be "full" and then we'll be hit with piles of constructed "urgent! crash landing!" pressure to increase again to prevent "full" regardless of the costs. E.g. a constant cycle of short term panic about an artificial condition pushing the system away from long term survivability.
I worry too, but you appear to be adding to this misunderstanding.
You are saying blocks can always be full (if the miners wished) because there are infinite 0 fee transactions in the memool.
Therefore, any increase is useless to get blocks less than full.
Therefore every increase will be filled with 0 fee transactions to infinity.
Blocks have recently been full of fee paying transactions, not 0 fee transactions.
Now adoption is on stop for a year.
(I know core want it this way, escalating fees. dynamic fee market)
0 fees is not an issue here.
0 fees have nothing to do with full blocks or required blocks size.