but the server's real-time clock was whacked and it still had a lot of drift. That raased problems with workstations connected to secure internet sites that had short time-outs - the clock would cause them to be logged out before they could complete the data input.
It was a hardware problem, but it took a while to get a replacement server box because mgmt didn't perceive it as a problem and replacement would cause an outage.
The two work-arounds we did were to decrease the server re-sync time to 30 minutes instead of 24 hours, so the clock couldn't drift too far; and set the workstations to sync to internet time directly, so their RTCs were correct in the "real world."
After two months, we finally got the new box, so it's moot at this point. But we got a real education about how much havoc a defective RTC can wreak.