Similar situation here. If all servers in my clusters suddenly showed a discrepancy that large, everyone would know, and I’d be scratching my head trying to troubleshoot the GPS timekeeper in the rack.