Thanks @HanOostdijk!
In a targets pipeline, if e.g., a CSV is defined as format = "file", then any change to the CSV is tracked and the entire pipeline downstream becomes out of date. A change can be as little as a single value in a cell.
I'm looking for a "best-practice" to work with a DB upstream in a targets pipeline, that makes sure changes to the DB table (like additions of new rows), are tracked. Much like dynamic datasets in targets, stored locally.