@cdn_yyz_yul I think that reading part is where your problem lies. When reading from same pipeline produced datasets, you shouldn't use 3 level name, instead follow DLT's way.

Older pipeline versions, where "target" field was used for target schema definition. You read like:

spark.read.table('LIVE.file1_table')

For newer DLT versions, where `schema` is used instead of `target` to define target schema, you read like:

spark.read.table('file1_table')

These are the DLT conventions to figure out dependencies correctly and process in order. If dependencies are not resolved properly, DLT processes them in parallel and hence downstream tables show now data.