In my perfect world, Fivetran would sync the column comments from the source postgres db and they'd get added to each of the column comments in Snowflake (or other DWHs). 

This would help reduce friction between engineering teams and analytics teams - no longer would we need to go to the engineers and ask what a specific column means or represents. That metadata would already exist within the table object.

Then, we'd be able to use the dbt_codegen package to auto-generate the source files along with their column descriptions. This would lead to us being able to dynamically generate the staging models AND the staging YML along with the metadata for each column.

My life would be so much easier. Please and thank you :)