As a guess, there could have been some software changes made that were a problem, but they were not propagated to all servers, leaving some bad and some good. Could also be interconnectivity failures between some servers and some databases... but not all. This sort of thing would account for some people having no issues and some getting routed to bad servers and having problems.