sipa and others have already explained it in detail on stack overflow. I am frankly tired of explaining it. go there to learn why the change was made

There’s a company that has a Bitcoin utility use case (not pics or vids) that can’t use cheaper witness data for some reason, and won’t fit in the 80 byte filtered OP_RETURN limit, and they really want to use the P2P network of nodes because their data is time-sensitive and they want to get it to as many miners as possible as fast as possible. So they are creating transactions that are technically valid with unspendable outputs because there’s no filter on that. But it means every node must carry all these unspendable UTXOs forever, which is bloaty and not nice. Whether you agree with this company or not, they found a way to do what they want to do while getting around current filters and they probably won’t be the last ones to do so. The hope is that this company or others that come along later with similar ambitions would use a bigger OP_RETURN instead, if that were an option. It is also possible that someone will come up with a use case for bigger OP_RETURN data that is less time sensitive. If lots of nodes are filtering those transactions they could go around the mempool and submit to large miners directly. IF this got valuable enough it could put smaller miners at a disadvantage to bigger miners in fees. If someone just wants to put arbitrary data on Bitcoin, it already costs much less to put it in witness data, so “spamming” OP_RETURN doesn’t make a lot of sense and if you REALLY want to, its technically valid right now anyway.