When is spam not spam?

My blog publishing software Movable Type has a fairly good comment spam filter, it's caught thirteen spam attempts in the last 24 hours. Every so often, however, one will fall through the net, and it'll be brought to my attention as an "unapproved" comment that I need to moderate. And even less frequently, a spam comment will end up being published as a real comment.

That's not a big deal, right? It just means I have to click the "mark comment as spam" button, and everything's cool, right? Well, technically, yes. But what if the spam is non-harmful? Most spam is a bunch of random words, and some URIs thrown in. But what if the spam is coherent, and non-linking?

Case in point: check out the latest comment on my References and Transcription entry, by one "Attassusa". It looks fairly innocuous. How do I know it's spam? I don't know for sure, but there are a number of factors: the comment is fairly vacuous; the email address provided is a free Russian webmail service; the comment arrived at the same time as other, more definitely spam-like, comments.

So the question is now, do I keep this comment, as it is harmless (as far as I can tell), or do I delete it out of principle? There's always the chance that this is a) a real person, with unlucky timing, or b) a test of my spam defences by spambots, and by letting it stay I've opened the floodgates for unmitigated assault. Any ideas?