(What makes this an unpopular opinion is that I'm basically advocating for jettisoning an entire Fediverse platform rather than just riding out the rough patch.)
I admin a Lemmy instance, and the amount of spam we get from Kbin and have to deal with is absolutely insane. Every morning I have to spend at least 15 minutes cleaning it up: responding to reports, banning accounts, purging posts, and pruning the images. And it's more of that all through the day, just in smaller bites.
Many times I've toyed with the idea of de-federating Kbin instances completely, but I don't want to cut off interactions with their legit users. The legit users are not the problem and are very much welcome.
Side note: If you're reading this and thinking "What Kbin spam?", then you should reach out to your instance admin(s) and give them a big 'thank you'. The spam from Kbin is not just a "me" problem.
The underlying issue is that Kbin mod actions do not federate to Lemmy. At all. The mods of their magazines can (and probably are) doing a great job cleaning up spam, but once that spam goes out, it's a one-way trip to Lemmy and completely up to the instance admins to deal with. For comparison, on Lemmy, the community mods can remove the spam which will then issue a removal action on instances that received it via federation.
I've tried various ways to unsubscribe and remove Kbin communities on my instance, but the "unsubscribe" doesn't seem to be honored by Kbin, either. So I'm stuck receiving content that's 90%+ spam and having to deal with it every day with no end in sight (barring de-federating Kbin completely). The only thing that sort-of works is hiding the Kbin magazines on my end, but that just hides the spam, too, and I'm still stuck storing it on my system.
So yeah, life for Lemmy admins would be easier if Kbin users would make the switch and I could just de-fed from Kbin without losing the good people.
What I find interesting that everyone just seems to argue that moderation actions should be federated out when the author claims that that's already how it "should" be (i.e. that's already the intent but it's not working). I never wrote code for either software and haven't even run my own instance or I'd try to reproduce the issue, because I suspect it isn't hard to pinpoint the problem if the fundamental code is already there.