Hi all, I’m trying to write an incremental transform that reads in the previous version of the dataset. I’ve followed the guidance in “maintaining pipeline performance” and added projections to improve the speed of reads, however we’re still running into slowdowns once the build has ran for awhile just due to sheer row count. Is there a way I can substitute in a View (that performs some PK deduplication/other compaction) for the “previous” version of the dataset without creating a circular dependency?