@rbtv It may execute something very similar to a `SELECT *` on the source table unless the platform adds its own partitioning or optimisation behind the scenes. From what I've observed, selecting batch snapshot often means the connector handles each run like a full snapshot load. The most straightforward solution would be to examine the generated rbtv query logs in Databricks or SQL Server Profiler. If no cursor column or incremental filter is specified, there typically isn't a smart `WHERE` clause to reduce reads.