There is a strictly limited supply of money available for spam. ( Because if we don't assume this, then we have to assume that all financial transactions will get crowded out by infinite funds. )

Every time OP_RETURN is used, instead of other UTXO bearing transactions, at least one UTXO is saved. UTXO spam is dangerous for decentralization (blockchain spam isn't since blocks are always the same size)

Since spam funds are limited, this is money spent to boost hashpower, and removes money from the spam pool that would otherwise create UTXO bloat.

Therefore it's irrational to prevent large (or at least multiple) being used for spam.

The only reason to limit this opcode is to "virtue signal that Bitcoin is not intended to store data".

And while this may have some benefits in the short term given the ongoing social attack against core maintainers, in the long term I think it might have very negative repercussions to cave to a social attack that has little technical merit.

However there is a serious issue with contiguous scriptlubkey data that doesn't have to do with government censorship.

Apparently it is possible to produce L2 protocols that rely on the relay network and unconfirmed transactions.

We need to be sure that minimum transaction fees are set quite high so that nodes are not relaying transactions that have no intention of being mined or paid for.