Fair point, yet I think it would be reasonable to think that the broad categories of techniques that allow storing arbitrary data on chain can be outlined.
I would imagine that everyone agrees that having to generate a trillion private keys until one is found that, when a bech32m address is created from it, has its first 5 characters after the human readable part be interpreted as the first bytes of some jpeg is quite more cumbersome than OP_FALSE OP_IF, right?
I believe not much would change in terms of actual transaction composition, in both mempools and blocks, even if v30 were to be broadly adopted - which does not seem the case to be honest - yet I cannot quite accept the idea that, since in principle spam cannot be prevented entirely, nothing should be done about it.