In medallion architecture we call this a "gold layer." Not sure how you're differentiating this through "rigidity" - common query patterns are a form of "rigidity".
You don't have to do this, the semantic layer will work anyhow for the full grain of the data model. You can just improve latency and reduce cost this way for very common grains.
> pre-aggregate data for common query patterns
In medallion architecture we call this a "gold layer." Not sure how you're differentiating this through "rigidity" - common query patterns are a form of "rigidity".
You don't have to do this, the semantic layer will work anyhow for the full grain of the data model. You can just improve latency and reduce cost this way for very common grains.