Personally I would try to avoid fetching descendants from the root node as that could potentially be a resource-demanding operation if the site structure is large. I guess it kind of comes down to how you've structured the multi-site solution. Personally I always use separate Document Types for each site/client and separate Indexes to ensure that one site/client's content never gets mixed in with another. If your site share the same types then I suppose you could look into extending the API with a custom selector that would filter the results based on the root node: https://docs.umbraco.com/umbraco-cms/v/13.latest-lts/reference/content-delivery-api/extension-api-for-querying