Tightening the consensus rules requires a soft fork, we are not there yet.

Consensus wise the OP_RETURN is already fully open. It was fully open in the relay rules by Core to prevent mining centralization.

No problem whatsoever, on the contrary, it helps prevent mining centralization:

https://njump.to/nevent1qqs8zeu9ufx8s3875rvu8w8x382jqg0at73rfhtjqw3tuwh03gyjj3gppemhxue69uhkummn9ekx7mp0qgsx9nhcswrrqg4y78tq64y90jwh99jswqkflc38kzvgczmwxmztensrqsqqqqqp0g4pud

No it is not, consensus rules can only be changed via a fork: a soft fork if you tighten the rules, and a hard fork if you loosen them.

Relay rules can be equal or tighter than consensus rules, but never looser.

The consensus rule for OP_RETURN size is that there's no limit. What Core is trying to do is make relay rules equal to the consensus rules (instead of tighter)

Why is this important?

When you have tighter relay rules, you create incentives for miners and pools to create an API where one can broadcast non standard transactions (which already happened) which increases the risk of mining centralization.

If I want to broadcast a non standard transaction, I just need to contact a handful of these APIs and send my transaction.

This makes a handful of big hashing power players to have a different mempool from the rest of the miners, a form of centralization, which creates opportunities for MEV and censorship.

And if you have censorship, it's not Bitcoin anymore.

So, you're worried with what people can put on the machine, while Core is worried with keeping the machine running.

Sorry to say, but in this specific case, Core worries are more important than yours: it doesn't matter what people put in the machine if the machine is not working.