The reported errors indicate that secondary validation has failed, which usual indicates some type of network geoblocking (of IPv6 in this case). The primary validations succeeded.
Also (and likely unrelated to the primary problem), I'm not sure why there are permanent, duplicate rewrites in your server blocks from certbot. That seems very odd to me. There are more than a few odd things in there to my eyes.
The v6 address is my own (from ARIN) 2602:f9ed:103::19 announced from AS21805, and there are no firewalls in the way.
I also do not believe it's any firewall as for the other domains, it picks up on the hard return 301 (URL), which is another issue I'll deal with separately.
The Nginx config is just perfect timing while certbot was running, to see what it looked like.
I'm still thinking this is related to some type of global reachability problem since the failures aren't seeing the HTTP->HTTPS redirect (because "http://" is in the address in the failures). Exactly what is returning the 404s I am not sure. It could be your nginx server, but that would seem odd given that the primary validations succeeded for the same content. I don't readily have servers configured to stimulate the multi-perspective global validation locations.
These results don't indicate much IMO, but they aren't great. They mean that your IPv4 and IPv6 responses differ, but they're not as conclusive as LE's global validation process. LE prefers IPv6 over IPv4, but will fallback to IPv4 if IPv6 isn't present, which IPv6 is present in your case.
Is there some type of caching/CDN present that might be interfering with the responses? Just throwing out ideas here. If you have other domains on the same webserver that succeed, this is possible. Also, but less likely, you could have orphaned nginx processes with a previous config still loaded that happen to (incorrectly) serve the challenge content.
That was it! I'm seriously scratching my head to figure out how that happened, because I did a full service nginx restart to be sure that wasn't the case, but checking the process list with ps confirmed it was exactly that.
Was nginx stopped when you did the certbot renew by any chance?
Because Certbot's --nginx option pretty much requires nginx to already be running. Certbot will actually start nginx but does not with systemd which can cause symptoms like the grief you saw.
Regarding the changes certbot makes, since it's one specific challenge per specific domain name (which is not as evident from the data sanitization you performed in your posts), there will be two locations and two rewrites, so all is well there.