If you can correlate semantically-parsed log data with time series to automatically understand that particular problem, you're solving problems harder than server monitoring and applying them to server monitoring, IMO :) Not that there's anything wrong with that!