I think generally when every alarm bell in your monitoring system goes off the first thing you do is question whether monitoring is broken. When you confirm there is a problem this big, you panic and try to fix it really fast. Then you call your other on-call guys and tell them you actually have an "oh, shit" situation.
Once you recognize there's a serious problem, THEN you make the public announcement. Ah, the life of ops.
They also don't seem to count the "partial degradation" periods as downtime; Stripe.js and Checkout.js both have a 9m red bar, but Checkout.js has had a lot of "partial degradation" periods. Both sit on 99.992% uptime, though. Maybe that's not so bad; it would depend on how severe a "partial degradation" can be and still be classified that way, I suppose.
Once you recognize there's a serious problem, THEN you make the public announcement. Ah, the life of ops.