The project team have said that the 2952 series is very chaotic with much more extreme runtime variability than other series. It seems that your task has a very chaotic parameter set. As long as the task doesn't hit the runtime limit for your system it should complete normally.
They cancelled the first batch of this series a couple of weeks ago (one of my cancelled tasks was at around 80% and had been running for 142 hours on my C2Q Q6600 @ 2.40GHz XP system), ran some checks on the parameter combinations and issued a new batch last week.
My C2D T7300 @ 2.00GHz Vista system has had runtimes ranging from 1 hour to almost 38 hours for the current batch of 2952 tasks and one user has reported a task running for 101 hours. The project team are investigating the parameter combinations for the current batch. I'd guess this is likely to be an iterative process, with a significant reduction in the number of upper end runtimes each time.
____________ "The ultimate test of a moral society is the kind of world that it leaves to its children." - Dietrich Bonhoeffer