If that delay reduces false alarms, when false alarms erode public trust in and concern about the alert system, then the massive benefit is clear. With such a system, drills and tests are going to be massively more common than actual events. Designing the UI so that the one-in-a-million option is as easy to invoke as the once-every-week option is a UI problem.
Notwithstanding it not being a primary issue in this particular case, I believe it's still an issue worth discussing.
Also, bear in mind that correcting the "deeper planning and procedural problems" would introduce delays as well. Anything but a purely automated system, or a big red button would introduce some delay.
No, it wouldn't, because those are problems that impact after an alert is issued, and whose existence caused delays in sending an all-clear. Resolving them won't effect the path prior to an alert, and would eliminate rather than introduce delays.
There was a confirmation screen.
I have seen no suggestion how a different confirmation than actually existed would have helped, and I don't think we want to introduce more delay in the critical path of a live alert without clear and massive benefit.
> A UI that makes it easy to make a mistake, even when the mistake is intended, and makes it almost impossible to correct that mistake, is a badly designed UI.
The problem with correcting the mistake is not a UI problem, but a deeper planning and procedure problem.