Bell curve.
Like a lot of things I think there's possibly an 'optimum' number of incidents, from which one can learn, however I'm less sure about accidents...
To my mind it's just like a normal bell curve - those at the upper end of the spectra, and those at the lower are people to examine for various [differing] reasons.
What could be interesting is to see if there's any correlation between the number of incidents v accidents for individuals. Some empirical data along these lines might then give some credence to some of the theories presented here...
FP.