Yesterday, once all SETI tasks were flushed, I checked my results list at the setiathome account page. There are 99 tasks in progress shown, all for the same host, all with download timestamps during the beginning of previous week's Friday outage due to campus router maintenance (but with different download times). I remember that this thing can happen when the server sends a scheduler response for new work, and this response appears successful to the server but is not actually seen or handled by the client.
(Confusingly, I don't recall that the respective host was restricted to only 100 - 99 = 1 task in progress afterwards. As some of you know, I had a slightly larger than usual count of hosts to supervise, but did so in a fairly organized manner, if I may say so myself.)
If it was a normal project, I would simply let these tasks time out. But with SETI@home's deadlines of cosmological proportions, I'll go and look in the setiathome forum whether there is a way to cancel such tasks after the fact.
Edit,
I was expecting to find an answer to this in a specific
sticky thread of the setiathome forum. And indeed there
is one in it. It's just highly annoying a) that they deleted the first post of the thread, making the info hard to find, and b) that the procedure is cargo cult. (The procedure
supposedly works, but nobody really knows what is going on. Or if there is anybody who actually knows it, they are incapable to explain it.)