Imagine a situation where a core fails to load because there's some kind of minor problem with the configuration for the core/collection.
Once you've fixed whatever caused the core to fail, there is no way provided in the admin UI to reload the core so it will start working. The CoreAdmin API can be accessed directly to initiate the RELOAD action, if the user is able to figure out how to do that. Restarting the entire node would take care of it, potentially with major disruption.
It would be really good to have cores that fail to initialize show up in the CoreAdmin section of the admin UI, in a different color and with some kind of visual indicator for the color blind, with a limited set of options that includes RELOAD.