Would also love any sort of communication back from Databricks on why create or replace view acts like this.
I have had to run my CICD once with create (or replace) and then make a code change to swap to "alter view," which per https://docs.databricks.com/aws/en/sql/language-manual/sql-ref-syntax-ddl-alter-view,
> is equivalent to a CREATE OR REPLACE VIEW statement on an existing view, except that privileges granted on the view are preserved.
It feels, sure...fine. But our table/view creates are run in the same workflow as our ETL, and making me deploy through my environments with something I can only run once to then change and deploy again so that I can change the guts of my view is not ideal.