The increased size is the same as what consensus enforces.
If you block it on relay level in P2P but allow in consensus, you make an incentive to submit txs outside of mempool. That breaks a few things such as fee estimation - you base the fees required based on what transactions are pending to be mined. If you can't see them, your fee estimation is wrong.
Relay policy and consensus policy should be the same, that works best for everyone and does not incentivize backroom deals with miners.
how would you escape homogeneity in that setup?
hmm, while (I think) I understand that logic, I don't understand why this supposed need of homogeneity overrides the need to have a blockchain as free of non-monetary data as possible (if that makes sense).
Shouldn't we have tried to limit the amount of data before letting it come to this point?
Again, honest question.
If you free it from non monetary data stored in one way, you don't go into a reality where there will be less non monetary data. You go into reality where it's stored in the worse way.
If people want to store data, they will (and they will pay for it). And they will probably use the cheapest way. So it's quite good, if the cheapest way was not bloating the chain state.
There is no reality without spam.
could you use the same rationale in an analogy to email spam from the perspective of someone hosting their own email account? This isn't some kind of trap; I'm curious how you would frame the analogy (with regards to spam filters, expected amount of spam emails and storing them, etc.)
I don't think it works for email. It would work if the sender paid fees to the receiver.
