For a continuously operating facility the question is never “when are you closed”. It is which systems tolerate which outage, for how long, and what has to remain available while they are down. That analysis happens in scoping and it drives everything downstream: the sequence, the crew size, the number of windows, and whether a parallel build is worth the additional hardware.
Redundant architectures are the friend of a relocation. Where there is an A side and a B side, the move becomes two smaller moves with the service continuously available across them. Failover is tested before the first leg rather than trusted, and the second leg does not start until the first has been validated and has run under load.
Where there is no redundancy, the outage is bounded by preparation. The destination rack is built, powered, cooled and cabled before the source is touched. Configuration is captured. The rollback point is written down with the condition that triggers it and the person authorised to call it. That decision is made in daylight, in a meeting, not at 3 a.m. by whoever is nearest the rack.
Month-end, quarter-end and reporting deadlines create their own freeze periods across the sector. Those are collected during scoping alongside the turnaround calendar, because a technically perfect window that lands on a production reporting deadline is not an available window.
