the old-school view of production control and scheduling is based on mainframe architecture - the application was hosted essentially on a single platform. if the application had to be up to match data entry shifts or jobs needed to execute to deliver reports by a certain time, then “do not touch” windows could be defined to minimize disruption.
newer architectures are harder to protect because the dependencies are difficult to discern. we can’t easily define where in the network user groups will relocate, and we don’t always recognize the relevant infrastructure components. That is unless we spend some money investing in, wait for it,,, configuration management.