It's the same thing as any other infectious disease, really. If you don't do any covid prevention and hang out with a bunch of unvaccinated people in close quarters, you're more likely to find out if somebody had covid, but at the cost of potentially infecting everyone.
This is about reducing harm by reducing infections.
Your argument boils down to "oh well, now there's CP forever accessible because its on this immutable data store, but at least we now know that somebody wanted to post CP".
That doesn't make any sense for CP. It's not like people who weren't into that crap before stumble upon it and are like "oh, now this is MY THING."
What I'd propose is having a "known bad content" (as flagged by users/operators) list that gives node operators a rating they can filter what they propagate by.
For example: if you have A, B, C, and D ratings (A being harmless, D being CP), you'd get a self-policing network because node operators could say "I only host/seed B+ content." Very few people given the choice would even want to seed D-rated content (and if I'm wrong, then we have a much bigger problem at a societal level).
It'd be imperfect, but so is blatant censorship. Arguably this approach would help you to find the worst stuff fastest.
What you are proposing is just censorship with extra steps, it seems to me, where the operators are the censors. What's the point, other than making everything more complicated and difficult to work with?
In essence. Except it's by consensus, not by decree. That's my personal issue with censorship. One or a few people deciding "this is bad," which starts out with good intentions eventually gets abused and used to subjugate groups of people. In this design, it's voluntary. You can be on the network, but that doesn't mean your stuff will be propagated.