I wonder if it would be possible to add a value that is the sum of all waiting times of all database accesses until the data is starting to arive. That way I would have a measure for connection lag + processing time and waiting time because of locks on the database or high amount of queries. That value per minute or something like this would show something is wrong like missing indices, complex queries or bad transactional design without going much into detail. I mean such a value should be not so complex to compute and should be suitable to see something is wrong. Of course it does not tell you much more than something is wrong. But looking at the other metrics helps then to dig deeper.