So you have a system. All you want to do is judge how the individuality of the components compares to certain norms. You determine what contributes to achieving the particular goal that the system needs to. And then judge how well each individual rates with the particular yard stick.
Lets complicate this a bit more. You have a bunch of these which do similar things. Each has its own yard stick. Each does its own matriculation. But they are all similar. The average judgment is what you would say, okay.
Suppose you introduce a freak in one of the system. This freak fails, all the tests that the yard stick is for. But ultimately can reach the goal in a more effective, albeit different way.
Now you should know. These systems are not all that simple. Each have a bunch of tasks to do and all of them are similar. For each task, there is a freak, but maybe is not very standoutish.
This yardstick, seems to rationalize all the tasks together.
Now, should the system be more robust to account for the freakishness of the freak? If you do rationalize, dont you think you are killing the individuality? But if you dont, the freak remains a freak. And the effective and better way of reaching the goal is lost.
In the middle of all this what is the freak to do? Apart from dealing with the nametag?