> Most plausibly, I would say, aeg managed to leave a process in an
> infinite loop.  Such processes can survive logouts by the user who
> started them.  983121.23 seconds is between 11 and 12 days; nothing at
> all implausible about that to me.  In short, I see no particular reason
> to think this is a bug in anything.

That would be plausibly, except that any such process only had 15 days
to accumulate 11+ days worth of CPU so it would have to be using 75%
of the CPU for 15 days.  I rather think I'd have noticed that :-)

I know that only 15 days were available as the only time aeg was
logged on during the accounting period was July 15 and 16.

--sjg