You are getting closer to a valid argument, but it remains incomplete. Filters still provide resistance to spam, making it more expensive the more nodes filter it. Your argument is essentially: "Since Core doesn’t filter spam in OP_RETURN or the UTXO set, and UTXOs are less efficient for the network, then spammers should be allowed to use OP_RETURN."
The real question is whether spam should be filtered. The only counterargument presented is "filters don’t work," which is a red herring. The argument against filters that do work is that they are centralizing, but this claim is made without evidence. Yes, IP RBLs are centralized, but they are an ancient and ineffective solution, especially in the era of Tor. Bayesian filters are far more effective, though they can be poisoned; again, an old problem. The fact that we don’t have perfect filters is not an argument against using filters at all.
The assertion that every mempool on every node must be identical is presented without justification, and that requirement itself is a centralizing force.
This leaves us with: "Core devs are dedicated and underpaid and shouldn’t have to argue with non-experts; they should just do what they’re good at, which is writing software." This is the most pathetic argument of all. It relies on argumentum ad authoritatum and argumentum ad passiones, and implies that we should entrust our future and fortune to people who are wholly socially inept. Not only is this a highly questionable request, it also leaves open the possibility that these same developers might be unable to defend their "technical" positions against more cunning and malevolent influences; whose presence we are fully aware of.