Catch-up prune for when nightly missed several days (e.g. after a restart). Same as run-nightly but max_rows defaults higher so the whole backlog gets cleared in one call. Use sparingly — this can briefly elevate write lock contention if the backlog is large.

JSON body (all optional): cutoff_date: default today (UTC) max_rows: default 10,000,000 (twice the normal cap)