Hi @532664 , I posted a "kind of" related post here. What I am doing is to save data from kafka as delta files, and this is my source of truth. Then I implement a medallion architecture from it, reading the raw data as cloudfiles from DLT pipeline.

In this case, as the raw data is correct, you just need to make a selective refresh to the bronze tables and its dependencies. I think that selective refresh isn't supported in continuous mode, but you can stop it, update to triggered, make the selective refresh, and then change to continuous again. It will be down for a while, but you can do that in a maintenance window.

I do not know if these are best practices or not. What do you think?