We have a KB https://support.elastic.co/knowledge/1f6ec6ed where we explain how to route data (esp related to APM flows) to different data streams.
I think it is valuable as "advanced use cases" section.
We should warn to avoid changing the dataset and the prefixes of data streams.
I think a recommended approach to separate data is to use the namespace of a data streams (<type>-<dataset>-<namespace>) because it's something we generally account in our Dashboards and curated UIs, while we are less tolerant to changes to dataset names.
We should also warn about the fact the variables/strings used as template must always be "valid" in terms of data stream/index naming for Elasticsearch (e.g. they should not contain - and be always lowercase).