There are a few different options, in no particular order:

1. Explicitly require normalisation.
2. Explicitly forbid normalisation (#993).
3. Make normalisation implementation-dependent.
4. Change the spec so that normalisation is no longer needed for most cases (#990).
5. Backout Unicode keys for now (#979).

(Am I forgetting any?)

I DO NOT want to talk about the content of the options here, but rather about how we can reach some form of consensus (preferably unanimously)? A few people have said things like "I would prefer X, but I'm okay with Y too". Okay, so maybe there's an option that can work for everyone?

Personally I'd be okay with options 1, 3, 4, and 5. That's actually all but one. I've made it pretty clear what my preferred path is and for what reasons, but I'm basically okay with most options except what's in main now.

Perhaps one option might actually be acceptable enough for an overwhelming majority?

There are still details that could (IMHO should) be addressed is we go with 1 or 3, and my proposal also needs a bit more work, but the main issue is the broad direction to go in – details are things we can discuss/change later, and are not necessarily highly contentious. It's just the broad direction that is.

So I propose everyone lists the options they find "acceptable enough", and let's see what overlap there is?

Or if anyone has a better idea? Ranked-choice voting? Something else?

I'm also okay with @pradyunsg just choosing an option (which may be an option I dislike) and we'll go with that. Hell, at this point I'd almost be okay with a random stranger off the street just choosing an option and just going with that.

Quite frankly I feel all of this is no longer worth my time. Things have become unpleasant and at times even silly. We're at about 292 comments in a quick count, and it's not just blocking the release of TOML 1.1, it's also chasing people away (as we've seen yesterday) and setting bad blood among regular contributors.

It's the most-discussed issue in the history of TOML by a considerable margin. All for what's really a minor feature that a relatively small number of people have asked for...