# Roman wrapper atlas-authoring status

Checkpoint: 2026-08-28, local build `20260828z`

## Decision

The statistical and authoring work needed to retire both broad Roman wrappers
is complete. Every bounded replacement is now authored into the public
playthrough and appears once. The old temporary `Broad wrapper retired` warning
is therefore deprecated in build `20260828z`.

This is a presentation dependency, not a remaining statistical-discovery gate.
The authoritative original supersession record is `roman_supersessions.json`;
the active immutable release is now `second_pass_v6_675`. The old direct URLs
remain available only as neutral, non-counted successor indices. They do not
enter playback and do not restore either wrapper as a statistical event.

## Current counts

| Retired wrapper | Current active v6 successors | Replacements authored once | Temporary warning deprecated? |
|---|---:|---:|---|
| `med-roman-republican-colonies-italy` | 57 | 57 | Yes |
| `med-roman-po-valley-colonization` | 15 | 15 | Yes |

The successors shared by both parent rosters count once statistically and once
in playback. No residual wrapper survives in the model or public case count.

## Notice-removal gate

The completed gate required every replacement assigned to each wrapper to have:

1. an atlas case or an explicitly adjudicated same-event node inside another
   atlas case;
2. reviewed prose, mechanism tags, affected/actor fields, and source links;
3. source-supported target, origin, route, and land-outcome geometry, including
   honest unknown or off-frame treatment where precision is unavailable;
4. chronology and any internal sequence authored without parent/child double
   counting;
5. its active v6 primary family and eight-trait card attached; and
6. exact-link, transition, wide, narrow, endpoint-land, and settled-frame QA.

All six conditions now pass. The broad wrappers are absent from the canonical
playthrough. Their stable URLs open index-level successor lists, which preserve
the explanation for old links without presenting a temporary warning or a
playable event.

## Evidence boundary

The completed research says the fifteen Po Valley replacements and the
fifty-seven Republican Italy replacements exhaust the recoverable event content
of their broad parents under the current scope. A later event-unit fairness
finding could still revise an individual child boundary, but no statistical,
source-review, coding, or public-authoring loop remains open merely because the
neutral old-link indices continue to exist.