My two machines typically put out 25k-27k of primegrid results a day... pretty steadily unless one of them gets the 3-second errors. Today & yesterday are ummm kinda special....
I'd guess that this is the cause:
http://www.primegrid.com/forum_thread.php?id=5197
right - that was the 2nd credit dump scenario i mentioned above - the stats sites would show zero credit earned for those days that the project server was down, even though your host may have been crunching work for that project during that downtime. so if the project was down for X days, then as soon as it goes back online, your host would dump X days worth of completed results onto the server in a single communication, which translates to a credit dump according to all the stats sites.I'm not sure how the database crash would affect people in a positive way... if anything I would expect that to cause a severe dip in numbers since work has basically gone up in smoke, as I understand it.
right - that was the 2nd credit dump scenario i mentioned above - the stats sites would show zero credit earned for those days that the project server was down, even though your host may have been crunching work for that project during that downtime. so if the project was down for X days, then as soon as it goes back online, your host would dump X days worth of completed results onto the server in a single communication, which translates to a credit dump according to all the stats sites.
that said, i think the only way such a project outage could be a negative thing for a host is if that host doesn't have enough work in the cache to last the duration of the outage. otherwise, all the credit not counted during the days of the outage ends up getting reported as a credit dump once the project goes back online...
It's probably the 3-second-errors crossed with the DB crash. http://www.primegrid.com/forum_thread.php?id=5196&nowrap=true#68082
Overall, it looks like invalid results are getting credit for others' valid work somehow. :\