Thank you for your reply!

Maybe I am slow, but two things don't sense to me with this reply. If these are clarified, I consider the question resolved.

You raise a good point that even if the you use AUTO CDC's sequence_by, you would still have the problem of reprocessing the same files if you don't have run_once on in the general case. In this case in particular though, this is not an issue as cloudFiles/AutoLoader keeps track of files it has already read. Smaller detail, but this is accurate?

For wider point of using a predefined schema, I guess that would work. However, then you miss out of schema evolution? There are pros and cons of enforcing schemas, and I guess it makes sense if you handle your CDC in silver that the schema is fixed by then, but I was otherwise thinking that it would evolve in bronze.

Or are you saying that, unlike other forms of pipelines in Databricks, a hidden downside of using declarative pipelines is that they don't support schema evolution? Does the declarative pipeline view handle schema evolution if the cdc files evolve? Let's leave schema evolution in AWS DMS as a topic for another day.