I find it disappointing how few people are addressing the proposed "Independent Tree Hash Headers" solution. It's probably the most interesting part of the article, but it's getting the least attention.

I came in expecting to disagree strongly with the article, but ended up agreeing more than I didn't (though I still don't 100% agree, as collisions are still an issue for mirrors). I find the concept of multiple hashes per commit quite interesting. It would allow mixing hashes in one repo, wouldn't break submodules, and tooling could be used to reject commits without any secure hashes for a gradual transition (like enforcing signed commits/tags).