Found out what's going on. There's a single SoB result with a ridiculously large cpu time of about 2.5 million hours. It's obviously inaccurate, for whatever reason, but that one task is throwing the whole average off.
Not sure yet if I'll tweak it, or just let it eventually get purged from the system in the normal course of events. I'll probably fix it, but it needs more thought first.
My lucky number is 75898524288+1