Had me an idea the other day that I finally acted on. I set my settings table antispam_last_update field to 2001-01-01 00:00:00 (which is the default value for "never updated") then emptied my antispam table. I then asked for the update and of course got the first 1000 published keywords. I then did my great experiment: I emptied the antispam table again. I then asked for the update 4 more times to get the rest of the published keywords.
Thus my antispam table is 1000 entries shorter than anyone else who happens to have the full list as of this exact moment. I will be undoing the little bits I do with my .htaccess file so's my blog is (effectively) wide open to referer spammers. Those would be the bit that searches referer for partial matches and the bit that says my comment post form has to be refered from my own domain. The only thing I'll have intact is my simple turing test for commenters.
My point here is to see if any of these old spammers are still active. In truth I already know (or strongly suspect) that quite a lot of them were NEVER active and were published by aggressive antispam administrators who didn't foresee a future with thousands of keywords and more added daily.
Anyway I figure I'll let it run this way for the month of April. Anyone else wanna join me in this test? Give it a month before reporting anything and we'll have a really good idea of what keywords down in the belly of the beast don't need to be keywords anymore.