So block size increase doesn't necessarily reduce fees paid per block, unless demand for block space is lower than block size. Ideally you want blocks to always be full. But a blockchain where new transactions are in the mempool waiting for a long time, if you double the block size, all the miner will do is add those in this block. It increases his income because he gets more transactions per block. The payout per block goes *up.* Fees go down so more people transact on chain. This is what dynamic block size in Monero is for, block size is only increased if it is profitable to do so, that way blocks stay full but can grow dynamically if demand increases. This is a queuing theory problem, you should look into queuing theory. The opposite can raise block fees, but can reduce profitability for miners, for the same reasons as above. Also, note that a primary reason for mining is processing transactions. Reducing block size doesn't make the network more useful and therefore valuable. Halving the block size doubles fees, doubling the block size halves fees, the miners make the same either way, but the utility is reduced when block sizes are smaller. It may be the case that it enables regular people to run nodes because the blockchain size increase is less, but again, unless you have to store the entire history of the chain to get 100% security guarantee, that's a non issue, a protocol like we talked about with MW doesn't need a block size at all.

It's actually not ideal for all transactions to get into blocks. Bidding balances times preference, so we get a equilibrium fee rate for different times. If the whole mempool goes into the next block, there's no reason to pay higher fees, so the miners lose. If we get to a point where all times have the same fees, then I can see a need for variable block sizes. Currently, you can cheap skate through if you don't mind waiting till the middle of the night for a north american.