There is a tempting moment at the end of a build: the deployment finishes, the page loads, and everyone calls it done.
The next morning is when the real system begins. Someone has to see whether the queue is stuck, restore a backup, understand a migration, rotate a certificate, or decide what the error means when the person who wrote the code is not online.
Done means operable
A production system needs logs and health endpoints that answer more than “is the process alive?” It needs a migration path, a backup that can be restored, and a response when an external dependency times out.
That is why the engineering notes put production inside development. The operational layer is part of what the architecture promises.
The infrastructure record is one example. Off-site backups and external monitoring are answers to the question “what happens if the host itself is the thing that failed?”
Launch proves that the path can exist. It does not prove that the path can survive a change, a restore, or a quiet failure.