Thankfully my memory is cloudy :-) maybe two points:
1. In general the idea of a separate reporting DB maintained by separate people who might not understand the semantics of schema changes seems inefficient in a world of cheap compute, compared to an application exposing key metrics directly. Upgrading every two weeks with an ETL pipeline off an app can result in RTL breaking every two weeks. Better just to release metrics from the same team that makes any other changes.
2. I think connecting hosted PowerBI to a IP-whitelisted Azure blob store is not possible, and that just seemed silly.
1. In general the idea of a separate reporting DB maintained by separate people who might not understand the semantics of schema changes seems inefficient in a world of cheap compute, compared to an application exposing key metrics directly. Upgrading every two weeks with an ETL pipeline off an app can result in RTL breaking every two weeks. Better just to release metrics from the same team that makes any other changes.
2. I think connecting hosted PowerBI to a IP-whitelisted Azure blob store is not possible, and that just seemed silly.