Got a source?

Not really. This is my own lived experience comparing Usenet to Google Groups, Reddit, web forums, and Wikipedia, and noting the explosion of user-contribution in the shift from Overcoming Bias to LessWrong. You could easily prove Usenet is declined, but I'm not sure what research you could do to prove that the incentives were structured wrong or that features like killfiles fostered complacency & reluctance to change, other than to note how all of Usenet's replacements were strikingly different from it in similar ways.

I don't think killfiles were a significant factor myself, but I admit I'm basing that opinion just on "it sounds wrong", not any actual data.

My read is that killfiles were a major aspect of systematically bad design of Usenet which made it uncompetitive and unscalable: it increased user costs it should not have, adding friction and trivial inconveniences. Killfiles express a fundamental contempt for user time: if there are 100 readers and 1 spammer, it should not take 100 reader actions to deal with the 1 spammers, as killfiles inherently tilt matters. What would be much better is if 10 readers take an action like downvoting and spare the remaining 90. Rinse and repeat. What is better, dealing with spam/trolls while using O(1) or O(n) in reader time?

I'd have attributed the decline of Usenet and mailing lists to (1) not being on the Web (that's the biggie) (2) barrier to entry to create a new discussion forum (even alt.* had process). Mostly (1) - the wine-users list (for Wine, the Windows compatibility layer for Linux) has a two-way gateway to a web forum, and immediately the forum was available the volume was 10x.

The non-Web thing is another example of this. Yes, an uber-nerd (and buybuydandavis is exemplifying this attitude in this thread) may contemptuously look at it as an irrelevant problem: 'what sort of person can't maintain a good killfile? or figure out how to deal with NNTP servers and ports and local clients?' But it's a big deal when repeatedly incurred by millions of people who do not wish to become uber-nerds and to whom costs matter.

Of course, all of this could have been fixed. But they weren't fixed in time, and so Usenet stagnated and died.