I can see now why using OP RETURN is less costly for nodes and safer than fake pubkeys or inscriptions. but i’m wondering, does this incremental limit 512 → 1024 really make a big difference? i mean, if an attacker really wants to, they could just pay the fees and send multiple OP RETURNs anyway also,
Currently there is a limit for one OP_RETURN output per transaction. In my "ideal scenario" I would preserve this limit for the next 2-3 Bitcoin versions. So the attacker (or NFT creator) would have to send several transactions to encode an image or video, and that would generate a lot of overhead for them.
It wouldn't make sense to develop a NFT protocol first for 512 bytes transactions, then 1024 and so on. At least, it would be very difficult to initiate a NFT fad (like the Ordinals fad in 2023/24) in this fashion. They would rather choose to use the existing fake public keys or the Taproot envelope format.
since some wallets or L2 solutions can batch or compress OP RETURN data the node impact can already be somewhat controlled does this limit actually change much?
Here I don't understand what you mean with L2s batching OP_RETURN data. But in general you are correct: the computational impact on nodes is low also if the limit is removed completely. My post was about the fears that a new NFT fad with thousands of junk transactions could develop right after the v30 publication. See also my conversation with @gmaxwell.