Must be talking about different studies - 607 RCDs tested well exceeds the Central Limit Theorem
Since you seem to be happy that the ESC's estimate of an in-service RCD failure rate of 2.8% is fairly accurate, I'd be interested in using this figure to gain some insight into your approach to 'mortal risk' - by presenting you with what I think is a reasonably close medical analogy ...
Imagine that it was known (e.g. because of family history or genetic tests) that you were at a small risk of developing a rare disease
(c.f. a small risk of suffering a electric shock potentially serious enough to kill) which, if you developed it, in the absence of prior treatment would probably kill you.
Having taken whatever measures are possible to prevent you developing that disease
(c.f. ensuring the basic electrical protection and ADS were in place), I then offer you some further options.
Firstly, I can give you a drug which, in the event of your developing the disease will have a 97.2% chance of preventing the disease killing you (i.e. a 2.8% chance of it failing to prevent your death).
Alternatively, I can give you another drug as well as the first, and this will increase the chance of your not dying if you develop the disease to, say, 99.6% (i.e. a 0.4% chance of it failing to prevent your death).
Neither drug has any particular downside (e.g. side effects) other than a small up-front monetary cost. If you chose the two-drug oprion, there would be a slight inconvenience in having to remember to take two drugs rather than one (or none)
(c.f. the slight inconvenience of maybe having to reset two RCDs if both trip)
What option do you choose:
1...No drug treatment, because the risk of the disease even developing is already small
(c.f. no RCDs at all, because the risk of your suffering a serious electric shock is already small)
2...Just the first drug – a risk of dying of 2.8% if you develop the disease
(c.f. one RCD)
3...Both drugs – a risk of dying of 0.4% if you develop the disease
(c.f. two RCDs)
?? Views will obviously vary - so I’d also be interested to know what readers other than you would choose.
Kind Regards, John