I don't think AMP would have a snowflakes chance in hell to succeed on it's own merits. It's success now is due to SERP-based strongarming by Google.
Given how simple it is load quickly, I think its inexplicable that AMP is so complex and its development so closed.
The only possible benefit I can see for Twitter to link to AMP is that their testing proves users prefer AMP pages, meaning a better user experience is result when "development is clearly steered by platform companies". When you say AMP is biased in the platform companies favour, what you really mean is it is biased in the user's favour.
As we've seen with the state of the non-amp mobile web, it's not simple to load quickly. Maybe it's simple to implement if that is your goal, but convincing all the involved stakeholders that it's a worthwhile goal is hard, and so the mobile web is slow as hell. The open web had their chance to make fast sites, and failed. So now an industry-led carrot-and-stick style approach is being used, and it's actually working. pages load fast now. And the HN Community, who is responsible for making a lot of those slow sites ("oh! let's load another three analytics platforms, it's only a couple megs") is angry.
Also, if the webmaster wants, they can simply add `rel=noreferrer` to the script tags. This is still considered valid by the AMP Validator.
Some ways of implementing fast (mobile) loading have the potential to be quite disruptive to Google's business model; AMP is a way to ensure that doesn't happen. Note that AMP is open-ish now, not when it was initially developed. The general model is thus a fait-accompli.
AMP is harmless enough nobody is likely to get all to worked up about it (despite these comments, neither am I!), but it does fit the mold of classic monopoly abuse rather neatly: using dominance in one sector to influence others in your favor.
Twitter, I can't see benefits, other than what you mention i.e. tests show users preferring it. My premise is that there is a large enough group of people who would prefer to opt out. I have no idea how large or small this group is.