Two other things we would generally recommend are 1.) Direct Connect and 2.) Using AWS DMS, or any other CDC tool set.  There are lots.  We have Query Federation now in the Unity Catalog, so you could set up the connection that way, but JDBC/ODBC connections are prone to errors / dropped connections, so for production data engineering tasks its better to load into S3 using a CDC tool first and use Databricks to load the data after.  If that's too heavy of a lift, then its pretty easy to use pyodbc and the sql server driver on a cluster once you setup the network access.