In our case , we process data in ADLS using databricks and copy into synapse. Users develop using QlikSense and PowerBI /SSRS. They are very comfortable with T-sql versus sparkSQL. I think over some time they will come to accept no T-sql and no SQL server but just delta lake.

We also opened up databricks-SQL to users and first issue found that it doesn't have event-based trigger to run the report. we can only schedule by time.

Also delta lake does not enforce referential integrity constraints on certain tables we have in sql server tables