The current code assumes that scoped regions are always subregions of those for region parameters. This is not true for nested functions like fn@().