This is one of those annoying safety paradoxes, subtly linked to the prevention paradox. The better you get at getting rid of the incidents the lower the rate,

This is one of those annoying safety paradoxes, subtly linked to the prevention paradox. The better you get at getting rid of the incidents the lower the rate, the longer it will take to gather data to improve further. There is a similar curve involved when it comes to debugging. High frequency bugs are easy to squash. Low frequency ones can be really hard.