I saw here that data expectation runs on the full output even when running incrementally. Does it run on the full input as well (also when just processing a single txn in incremental mode).
https://palantir.com/docs/foundry/maintaining-pipelines/define-data-expectations#incremental