Again, this seems straightforward rather than problematic. It’s unreasonable to expect anyone to issue a configuration command and then look at a screen for 10 minutes until it decides whether this is good or not. So the problem is sorted: it can’t wait without it becoming a terrible user experience. Anything else is fine: do whatever pre-checks it can do to validate the configuration, wait for a short little while, spawn a process that waits in the background until its time to apply the configuration, provide a command that checks whether the configuration is good, etc.
In the future we can also provide recurring health-checks which allow the service to report issues once it’s able to, but that won’t happen today or tomorrow.