Looking good! I think this is much better than the original draft. Agree with Andreas that supports_instant is simply equal to (supported_instant_providers.size() > 1) which makes it redundant.

Daniel is right that putting every possible provider in the Payment message might not scale in a world where there are huge numbers of instant-confirmation providers, but I'm hoping that we never have to scale to that size, because if we did that'd rather imply that Bitcoin has become useless for in-person payments without trusted third parties and avoiding that is rather the whole point of the project :) So I'm OK with some theoretical lack of scalability for now.

A more scalable approach would be for the user to send the name and signature of their "instant provider" every time and the merchant just chooses whether to ignore it or not, but as Lawrence points out, this is incompatible with the provider charging extra fees for "instantness". But should we care? I'm trying to imagine what the purchasing experience is like with optional paid-for third party anti-double-spend protection. Ultimately it's the merchant who cares about this, not me, so why would I ever pay? It makes no sense for me to pay for double spend protection for the merchant: after all, I'm honest. This is why it doesn't make sense for me to pay miners fees either, it's the receiver who cares about confirmations, not the sender.

So I wonder if a smarter design is that the user always submits the details of their instantness provider and we just don't allow for optional selection of instantness. I'm not sure that works, UX wise, so is having a less scalable design to support it worthwhile?