Another question: if I note the average of latches goes up, what actions I need to take to resolve it? (if it is really a problem as indicated nagios).
The only 'normal' that counter has is 'normal for your system'. It's dependant on hardware, database (structure and size), workload and probably other factors.