Which destination?:

Problem:
When a column type changes in a MySQL source 
(e.g. INT → BIGINT), Fivetran performs a 
CREATE OR REPLACE TABLE AS SELECT (CTAS) on 
the Databricks Delta destination table instead 
of an in-place schema evolution. This completely 
wipes all Change Data Feed (CDF) history on the 
affected table.

Environment:
  Source          : Aurora MySQL 3.x
  Destination     : Databricks Unity Catalog
  HVA version     : 6.2.0/45
  Proxy Agent     : 1.2.3
  Databricks DBR  : 18.0.x
  Delta Sharing   : No

What we tested:
  We enabled delta.enableTypeWidening=true 
  on affected tables but Fivetran still 
  triggered CTAS. The property had no effect 
  on Fivetran's behavior.

Expected behavior:
  Fivetran should handle column type widening 
  (e.g. INT → BIGINT) using Databricks native 
  ALTER TABLE ... ALTER COLUMN TYPE or 
  delta.enableTypeWidening without recreating 
  the table, so CDF history is preserved.

Impact:
  CDF history is the foundation of our 
  historical reporting and SCD2 views. 
  Every column type change in MySQL permanently 
  wipes CDF history on that table breaking 
  all downstream historical queries.
  We are now fully live on prod with this 
  limitation as a known risk.

Support ticket reference: #396359