This grid is a better one to look at. SpamAssassin 4.0.1 test grid has other things going on. The Acme::TaintTest module I wrote is supposed to isolate just this one problem. Since it only happens on certain test machines, I have to upload a test and then wait for as many of the relevant test machines as possible to happen to pick it up. As of the time I'm writing this, that grid for Acme-TaintTest 0.0.8-rc1b-TRIAL does show the taint error in some, but not all or most, of BINGOS and Slaven Rezić's machines. I tried reaching out to them when all I knew about this was what I could glean from the SpamAssassin test failures, but that didn't go anywhere. I'm hoping to narrow things down much more with the Acme::TaintTest series of tests and the conversation here before I reach out to them again.