Hi @zychoo ,

I would consider a “Near-real-time” solution into Databricks Bronze, something like:

- Log-based CDC tool (Qlik / Debezium / HVR) captures changes from SQL Server.
- Tool serializes sql_variant to JSON or string+type metadata.
- Writes to S3/Blob as Delta-friendly format (JSON or Parquet).
- Databricks Auto Loader streams into Bronze tables.
- Silver layer casts back values using the type metadata if needed.

This avoids snapshots + fragile CDC merge logic in ADF.