I’m very new to ebt, however I suspect you have to create a staging-like table for the incremental data in the csv, then run a separate stg to run a union referencing the csv data that has been materialized in the underlying db. I say this because the seed capability is a specialized capability only meant for small data imports of standalone tables. I also say this because even using the db directly often requires running a union of a temp table that stages the import.
FYI: The following link talks about using a macro instead of a seed.