I don't follow how PoS allows Steem to process more transactions.
The ability to produce a candidate block, propagate it to the network efficiently, store it on a server and access the data is quite separate from the consensus method of creating or validating the block header (such as Proof of Work, Proof of Stake, Delegated Proof of Stake) which allows it to become part of the blockchain.
Correct me if I'm wrong, but would it not be possible to modify Graphene to create a PoW blockchain that can process the same number of transactions as the Steem blockchain? The only issue I can see arising from this is a high rate of orphaning, but blockchains can function pretty well even with high orphaning rates, it's just a frustration for miners.
My understanding of the benefit of Delegated Proof of Stake is that it functions essentially as the same consensus model as Proof of Work but makes the system economically more efficient (ie. as Bitcoin gets more expensive, they necessarily must increase the amount of energy burnt every day until miners are getting normal but not supernormal profit), rather than efficient at the level of nodes and transaction processing. That energy represents spent capital for the system as a whole, a cost avoided by DPoS while retaining the same incentives (trying to cheat in block production incurs risk of financial loss in both models).
TLDR: Proof of Work is wasteful, but it doesn't limit how many transactions a blockchain network can process. The limits imposed in Bitcoin and others including Ethereum are arbitrary values with the goal of limiting "centralization" due to high cost of operating a non-mining node.
My point is that DPoS is an integral part of a system that can process that many transactions. You take away DPoS and everything changes and no, you can't say that if you removed DPoS everything would work exactly the same and the network would process just as many transactions, especially if you replace it with PoW. The network would take longer to come to consensus. The amount of energy consumed by the network in order to achieve consensus would be orders of magnitude larger and that cost would have to be paid by someone through fees. You're saying that simply because you can not draw a causal line from DPoS to more transactions, that DPoS is not integral to a system that is capable of processing that many transactions. Of course that begs the question, "Then why are no other blockchains running PoW and capable of processing that many transactions?" DPoS has many consequences within the system. Even just the fact that it enables more frequent hardforks is a factor because that enables the community to make changes to the platform that help improve performance.
The time it takes for a network to come to consensus is the time it takes for a block to be propagated and validated across the network. That is a function of block size and network latency/bandwidth.
http://www.tik.ee.ethz.ch/file/49318d3f56c1d525aabf7fda78b23fc0/P2P2013_041.pdf
Block time in general is a decision made as part of the protocol development, weighing the value of granularity of information vs. the fears of centralization where network latency gives an advantage to larger players working together. Ethereum and others could have chosen faster block times from a technical perspective, but weighed them up in this way.
https://blog.ethereum.org/2015/09/14/on-slow-and-fast-block-times/
In my estimation, they are just fears. Every time block time has been reduced or block sizes have been increased they have been found to work well without serious drawbacks.
I can't answer that with certainty but my view is that it is a matter of fear. There is a certain culture pervasive in the crypto community which does not seem to be or have been present in the development of Steem. Steem has firmly embraced the idea of nodes being run by institutions from the start, not held back by the expectation that end users must run full nodes to validate everything locally. Most other cryptocurrencies have inherited that from the Bitcoin community. Even Bitcoin Cash has maintained 8MB limits (despite successfully testing 1GB blocks) out of conservatism.
I agree with your points on the financial costs added to the system and the comparative ease of hard forks allowing faster development progress.
It seems Bitcoin Cash (and all associated that are now cooperating) are considering unlimited blocks, but if I understand it correctly there is still a risk (a fear and maybe just a fear, I'm not certain) that the limit is needed to avoid an attack.
I used to be sceptical of any Bitcoin forks, but as of late my respect have grown for them. The Bitcoin forks that is, not for anyone in particular that might be working on or supporting them.