Mickey is correct that the need for weekly testing comes from the fact that the weekly fire alarm test does not do the job. Let me have one last try.
See a green break glass is normally there in case the fire alarm interface fails. Probability very low (call it x where x is a very small fraction, like 1 in 100,000), consequences kind of pretty bad, especially if there are people who are not familiar with the normal means of egress control. If the fire alarm system fails you rely on the normal mushroom head push button (if present). It goes through an access control system commonly which is not designed as a life safety system (and I have personally known them to fail and had to deal with fall out), but probability y (y low but nothing like as low as x). So we have the green break glass. Probability of failure very low (a bit like x, but call it z).
In a fire alarm system, because z is low and the consequences of failure (someone has to use another call point to raise the alarm) may not be too awful, we live with annual testing.
In a conventional arrangement, for people to be trapped in a burning building, the probability is xyz, which is so low you can forget about it, particularly as the weekly fire alarm test means that the down time of the very low probability event is short.
Now look at Davey's wee building. If there is no fire alarm interface, probability of people being trapped is much greater, namely yz if you have the mushroom heads or just z if not. So for example, say x were 1 in 100,000, Davey's trapped people are more likely to die by a factor of 100,000 than people in conventional arrangement. If a door is even semi critical, why would you, as well as massively increasing the likelihood of failure, increase the down time for which it is allowed to exist. Microswitches do stick (though how I know this since all I do is sit at a desk and write standards I am not sure).
Cant see it myself but not sure I really care.
