Inconsistent Behavior When Using Lakehouse Tables in Warehouse-Scoped Queries in Copy Activity

Hi Fabric Team,

I’m writing to report a bug we've encountered while using Microsoft Fabric. we rely on Fabric for our ETL and ELT processes, leveraging both Warehouse and Lakehouse features.

As you know, when setting up a copy activity, we are typically allowed to use a query only on Warehouse, not on Lakehouse tables. However, we discovered that Fabric currently allows users to reference Lakehouse tables in a Warehouse-scoped query, which results in subtle and hard-to-detect issues.

This behavior caused a critical problem in one of our production pipelines: records from yesterday’s load were not copied from Fabric to Oracle. After thorough investigation, we found that the issue stemmed from a query referencing a Lakehouse table, even though the copy activity was configured to use a Warehouse. Fabric did not validate this inconsistency, and as a result, the copy activity silently failed to move the expected data.

This bug could potentially have significant consequences in production environments. I recommend implementing a validation step to ensure that all objects used in a query are consistent with the selected source type (Warehouse vs. Lakehouse), to avoid such hidden issues in the future.

Kind regards,
Yazdan
Data Engineer