Move "Prevent updates to the primary" to be the T minus zero point
When we are doing the rehearsals, we start the call and do some Pre-flight checks, before we commence with Phase 1: Block non-essential network access to the primary. According to the notes this usually takes between 5 and 20 minutes.
But on D-day that would mean we announce GitLab will go down from e.g. 13:00 UTC, but in fact it will only be down 5 to 20 minutes later. So I'd like to suggest we start the call 15 to 30min before the blackout window, so we can start Phase 1 exactly at the time we announced the blackout window would be.
For absolute downtime, this doesn't make a difference, because it does not speed up the failover process. But for our users, they see the downtime window for the announce start time until it's back. If we cut off the uncertain period (time between announced start time and when users actually can no longer access the site), we'll be up shorter after the announced start time.
I can file a MR to update the failover template, but before I do, I wanted your opinion on this. @dawsmith @glopezfernandez @nick.thomas