Is it just a question of variance in the size of the issue? Is having 10 incidents of 10 people each the same as one with 100?

I’ve argued in systems reliability that once you are already high reliability you are better off lowering incident size vs total unavailability.