What would have to be changed in 'inn' to make it better than what it already does other than possibly auto-peer finding?
Personally, I think auto-peering would be terrible for the reason I mentioned in my previous post. If you have to manually set up a peer, you have some control over who and what you propagate. If your friend starts posting garbage, one of your other peers might get pissed off enough to de-peer you--which I consider a feature! Configuring a peer isn't difficult, but it does require an explicit human-to-human agreement.
The way things work today, when somebody wants to post on Usenet, they go to eternal-september.org and get a free account. I think something like 50% of postings on text newsgroups must originate from eternal-september, and it's great that they're providing a free service, but things like e-s and the Google Groups Usenet bridge means that there's no accountability for garbage. With a $20 Raspberry Pi, though, you can run a totally capable text groups server, and I think a network like that might have some interesting people on it. It's also a foot in the door toward people hosting their own content rather than shoving everything on some random free online service.
What Mastodon does for decentralized / less-shitty Twitter, this could do for Usenet.
Edit: I'm going to go ahead and put it out there: people who are seriously interested in doing something like this, there's a link to contact info in my HN profile. Email me. A dozen users from HN plus users from the more curmudgeonly online communities I follow may almost be enough to make it happen.
Most Usenet clients would filter on the overview headers, which has the advantage that the messages did not have to be downloaded first, so it's much faster (these days that might not be such a issue), but the disadvantage that all a spammer/troll has to do (and did do) was change the From address to bypass the filter.
With Nfilter it was possible to filter on anything, for example crossposts (commonplace with trolls and spammers), NNTP-Posting-Host or even the Path, so a bad Usenet site with nothing but spam or trolls could be completely filtered out or flagged as junk.
The problem is nobody wanted to make even the small effort of running Nfilter (I tried repeatedly to get people to use Nfilter when groups were invaded by trolls but they preferred to just mindlessly continue complaining about it instead).
The author of Nfilter gave up development in the early 2000s, but its site is on Wayback Machine, from where it can be downloaded from its archived page (whether it works on anything later than Windows 95/NT I don't know), along with its source code:
So even without individual users running INN (which seems a bit like overkill to me), users can and always could filter out every type of abusive garbage very easily. They just have to want to do it.
The Usenet Death Penalty (UDP) is what was used in the past against ISPs or spammers, but the wikipedia article  indicates that it relied on cancel messages rather than reconfiguring the peering arrangement.