An audit log should let the public look backward. But eight overlapping exports from West Des Moines show a narrower view: whole rows turn over between snapshots in a pattern consistent with a rolling window of roughly 30 days. [F-027]
The analysis found no row-level field mutation. It does not establish that anyone changed a record, and it does not identify an actor, mechanism, or technical cause. [F-027]
A moving window, not a stable ledger
When an export behaves like a moving window, obtaining a file on one date may not preserve what was visible in an earlier file. That limits the ability to reconstruct a long-term history from later productions alone. [F-027]
The technical source of the turnover is not established. It could originate in the underlying system, the export process, or the display layer. The available evidence cannot distinguish among those possibilities. [F-027]
That uncertainty does not erase the public-records problem. Oversight depends on records that can be preserved, compared, and understood over time. A requester cannot audit what a later export no longer contains unless an earlier copy was separately retained.
A correction built into the finding
An earlier derived spreadsheet described 39 “swap pairs.” That count does not reproduce from the raw exports under independent re-analysis and is not relied upon. The reproducible finding is whole-row turnover consistent with the rolling window—not changes within individual rows. [F-027]
An independent analysis reached the same overall rolling-window pattern by a separate method. It also flagged 17 records reappearing non-sequentially across snapshots, a feature the rolling-window explanation alone does not fully resolve. A separate per-day mutation-rate figure from that analysis did not reproduce and is not relied upon. [F-027]