In Java, Redis client has a structured API, but then emits a query syntax. The Java instrumentation is written at the lower-level query layer and so captures db.query.text. However, if the instrumentation was written at the structured API layer, then I think it wouldn't capture db.query.text.

This could mean a lack of consistency between Redis instrumentations with respect to db.query.text, but there's already inconsistency with respect to db.query.text since it should only be captured if the instrumentation implements sanitization (which not all instrumentations will), so maybe it's ok.