We had confirmation from Databricks that the approach detailed in my reply was the only workable approach for user-based SSO. We ended up replicating data across from Snowflake into Databricks in the end. Another approach is to use a service account etc, however this might break some data governance and audit policies if your org. requires such things.

Something else to consider is that if your query can be pushed down into Snowflake then it gets executed on the Snowflake side and the results returned via the connector. As soon as a user executes a query in Databricks that can't be pushed down (e.g., uses a UDF) then the data for every table referenced in that query is pulled through to Databricks and the query is executed on the Databricks side. This will have a very significant impact on query performance, therefore depending on your use-case, data volume, and frequency of Snowflake being queried you might want to consider replication.

https://docs.snowflake.com/en/user-guide/spark-connector-use#pushdown