I've put something together using GitHub Pages; the advantage being that the metadata can stay in a machine-readable format (in this case, YAML), while the Jekyll engine can render pretty pages and do calculations for human consumption. (The choice of technologies is maybe not my immediate preference, but this strikes me as most readily sustainable, by far.)
I intentionally kept the data a bit leaner, though, than Brian's original suggestions, and have backfilled metadata through changes the tz mailing list was first informed of since 2016-01-01T00Z. 2016 had a lot of changes (particularly in Russia) which were going on simultaneously, which made tracking down some of the original emails more difficult, and so there's a bit of guesswork involved. I don't intend to go back any further than this, nor do I particularly encourage others to do so, either.
If there are glaring omissions that keep this resource from being as useful as it could be, I'm happy to hear it, though I am interested in keeping this relatively lean so that it can remain fairly easily updated as new changes to tzdata arise.