I'm intrigued. Would you consider this an example of the type of failure that you envisage.

As computers increasingly become embedded in sensors and actuators, some types of device, notably medical measurement/intervention devices and smart maters, will change behaviour over time. This is an opportunity that should be exploited (new insights can be rapidly exploited/risks avoided), but existing models of validating devices based on a one off evaluation will fail.

I've seen this not picked up well by the MHRA, which has validation cycles designed for hardware defined devices that take many iterations of a software defined device to execute. Also, Ofgem/OPSS seem to have gaps in who is picking up smart meter failures that may be difficult to spot for consumers, and which may have a significant effect on vulnerable consumers.