Rotate sampling records for data scientists before hourly run

Your sampling record flags pile up and skew model validation, causing missed rotations and stale test sets. This delivers one inspected record at a time so validation stays current before the next hourly run.