The devil is in the details. How "wrong" censoring spam is really depends on how it done. How much decentralization was given up in the process? How many legitimate users were silenced inadvertently? This is a delicate issue, if it's to be handled optimally.
Oh, no argument--it's a VERY complex issue--and "filtering" is one solution-- Yet it relies on counting on relay operators to then not censor other things...and that's when it gets sticky... It's really the exact problem Facebook, Twitter, etc. are faced with--how much censorship is "ok"? And who gets to decide?