- In the meantime, Bitcoin -- at 1MB blocks -- is hard-capped at ~250,000 transactions per day. Period. The consequences of this could rationally be argued to be constraining new entrants into the system. (Indeed, I would say it is obvious that it does, but grant that others may have a differing opinion.) The pool of node operators is a self-selected subset of Bitcoiners. If we prevent new Bitcoiners from entering the system, the potential population from which node operators are selected cannot grow. Many of us think it likely that some percentage of new entrants will start up nodes. So increasing block size may actually result in more operational full nodes.
This is a key point. More people being able to use the bitcoin network will be a decentralizing force. Really, the debate is about which force is stronger... nodes dropping out from extra bandwidth or new nodes because of new users. If the percentage of users running a full node drops from 5% to 2.5%, but we double our user base, we have the same level of decentralization. Since increase in bandwidth and storage are generally better than linear, I would expect that increasing usage and thus resource requirements would result in a sublinear increase in node operations costs. If you can have X bandwidth for $Y, you can probably have 10X bandwidth for $2Y.
You might argue that increasing blocksize will result in more transactions of less value, but changing blocksize has 0 influence on the miners' cost to include the transaction in a block. There is always a cost due to orphan risk, long term storage, and UTXO bloat. Blocksize doesn't change any of that.