In the offered example, the problem was that the mounting bolts apparently were not torqued during original installation and apparently were not checked during periodic maintenance, and progressively crept out under sustained load plus thermal cycling and vibration. All of the real-world-plausible mitigations for the GE design would have been to likelihood, not severity.
Back at original design time, severity could have been mitigated by designing the system so that the gamma camera was lateral to the patient, who would have been laying "on edge". Then the severity (of harm to the patient) of the camera falling off its mount would be minor. i.e. it would delay the medical imaging procedure or cause it to be re-located to another imaging suite. Such a camera configuration however would be medically inferior in regard to skeletal alignment and gravitational organ displacement. So instead, apparently, mitigation of likelihood was accepted as the better approach. The problem was that follow-through on that likelihood mitigation was not properly implemented.
If the point of this particular example was to argue that severity cannot be reduced, it's not a good example.
Once again, it comes down to how things are defined, and there are more than one "right" way to do it.
Then the severity (of harm to the patient) of the camera falling off its mount would be minor.
I disagree with this approach. "Camera falling off its mount" is not a harm IMO. Patient death (or injury) is a harm; Camera loss is a harm; Prevention of timely diagnosis is a harm.
The severity of patient death (as unlikely as it may be) in the hazardous situation of a camera falling off its mount right next to the patient is still top. The probability of this hazardous situation leading to this specific harm is the big risk difference between top mounting and lateral mounting.
Cheers,
Ronen.