I philosophically disagree with the idea of insulating the schema from config changes; the whole point of having a dynamic schema is that we should be able to handle such changes with relative ease. There is admittedly a problem with linked configuration options, but that is a more general one I don't want to workaround by doing damage to the schema.
That said, I'm a bit less concerned about this particular option. That's partly because setting the reserved flag to False for all stars is a natural limit as reserveFraction goes to zero (i.e. the presence of the field isn't confusing when the config option is set to a value that doesn't require it), and partly because the cost of the empty flag is just one bit per source. If either of those wasn't true, I'd fight a lot harder here.