What's the largest contributing factor to increased propagation delay? Is it transaction verification, a bandwidth issue, maybe something else? Verifying more transactions would slow down block creation, but I don't think it would by too much. As for bandwidth, miners don't send out blocks to every node, they send them to their peers who then propagate them further, so I'm not quite sure I understand how bandwidth is a problem since there are plenty of nodes transmitting and receiving data.
Bigger blocks take longer to propagate to the network and they also take longer to verify. The latter is especially problematic at larger block sizes because for certain transactions,
signature hashing quadratically scales. If it takes several minutes to verify a block, that's a real problem. If invalid, it'll be orphaned. If miners are doing validationless mining to sidestep the issue, this can result in orphaned chains. This means less secure transactions for users, and more expensive mining for miners. This encourages smaller miners to shut down and larger miners to scale up to account for these inefficiencies.
Segwit partially solved the quadratic sig hashing issue, but transaction data in the base block is still a problem.
Without inflation, the system needs fee revenue to continue incentivizing miners. The block size limit is the only means we have to enforce scarcity of block space, which guarantees fee revenue. Otherwise, Bitcoin's Byzantine fault tolerance may be threatened as block rewards decline in value. You can't have a network worth many billions or trillions of USD where miners have no incentive to secure the network.
Agreed, but why would users want to pay massive fees to cover the miners expenses, especially if they can use other networks for far cheaper (I know I know "they get a secure decentralized network" but still, does this limit the types of transactions that will be capable on the bitcoin network? Is this where LN comes in?). Is it fair to assume that up til now (and maybe for a few more years) miners have been taking the hit in terms of fees with the belief that their accumulated BTC will be worth more in the future than their current mining operation has cost them (i.e a net positive in the future)?
That's a fair assumption. Miners are paying far more per transaction than users are. Rationally, this is because they are investors who are speculating on the future value of bitcoins. It's bad to assume that miner speculation on fiat value will sustain the system forever, especially if the block rewards get smaller and smaller due to lack of fee revenue. At some point in the future, Bitcoin's value should become less speculative (as it becomes an established asset class) and mining incentives should accordingly be more in line with actual usage. Mining can't remain entirely dependent on speculation. That was never Bitcoin's design and it's just not sustainable.
If Bitcoin is a settlement layer, it will probably limit on-chain transactions to those who need to transfer higher value, or those who have a specific need for Bitcoin confirmation -- like a Lightning commitment transaction.
If people aren't willing to pay high fees, then mining revenue will drop and mining security will therefore drop. This may be fine, and we will hopefully see an equilibrium form where mining security reflects what users are actually willing to pay.