This may have nothing to do with indexed data corruption. If you had something that could generate 1,000 log events in a second, and if your timestamp is accurate only to the second, then all of those log entries would have the same epoch time (which is given in the error message for the frame of time where > 1,000,000 events have been seen). The epoch time is accurate to the millisecond, so in a single second you have 1,000 epoch times but only one is being used for all the events begin logged in any given second. Windows iis logs are by default accurate to the second. But you can google for your version and see about turning on millisecond time stamp. Whatever logs you are using, that's what I'd try.
If this is being caused by corrupted data, there is quite a bit of information about dealing with that here: http://wiki.splunk.com/Community:PostCrashFsckRepair
