Exactly. It's super evident to anyone who works on maintaining SaaS that is in production with at least a hundred users at any given time using it in real-time when you are trying to push features or fixes that even going down for a minute will result in serious service reputation damage. We engineer a lot of reusable scaffolding and reporting tools to modify our services in-situ with minimal/no disruption because it is a huge priority. Operating on public infrastructure is the closest real-world parallel.

But to most people who work jobs where their choices don't directly influence customer service uptime (including a construction worker or a surveyor of construction for the same public works, as opposed to the project coordinator/planner/senior architect/engineer), this is an opaque cost.