# L4RA scene and placement audit

The viewer uses a single installed `STREAML4RA.BUN` for its default Bayview
free-roam scene. The companion `L4RA.BUN` has an explicit `FREE_ROAM` group;
the matching free-roam route export contains 832 route records. The source
executable constructs one `TRACKS\STREAM%s.BUN` path from the active track
record. The available static trace does not establish that L4RA alone contains
every retail city area, and it provides no source-backed plan to join all eight
L4R streams as adjacent map tiles.

The current exporter removes scenery placements that are exclusively members
of numeric race barrier groups. The [event-group audit](../../reverse-engineering/city-event-group-filter-r397.md)
records the table crosswalk and counts. This filter does not adjust the
positions of any retained building or prop.

The latest [eight-scene browser review](images/539-r398-all-regions-contact.png)
shows the limits directly: the scenes overlap or represent distinct track
contexts, and two contain no free-roam route data. The [stream overlap audit](../../reverse-engineering/city-region-overlap-audit-r203.md)
records the executable and geometry evidence. The combined eight-stream union
remains experimental because it introduces mismatched scenery and road
intersections.

The [current L4RA drive capture](images/530-r397-l4ra-event-groups.png) loaded
all 223 nearby textures in host Chrome without page errors. It shows a usable
street view, but does not establish correct placement throughout the city,
retail lighting, original handling, or completeness. The XP guest GPU is not
used to render this export.

## Latest local road/building checks · r403

A fresh packaged Chrome load at Route 450, distance 525, radius 160 reached
`City scene ready` with 89,205 nearby triangles, 223/223 uploaded textures,
832 L4RA routes, and 12 ambient body models. The player start was at
`(-867.00, -444.14, 21.59)`. The debug triangle-intersection query returned
zero placed-object/road intersections within 100 world units of that start.
This is one local crop; it does not certify the rest of L4RA.

The older Route 530 `XB_CASINOD_SIDEA_1A_LL_00` overlap also should not be
treated as a placement error. The triangle audit found only 44 facade-edge
crossings near the road boundary; clipping them opened gaps in the wall, while
the unmodified passage stayed drivable in the viewer. See the
[Route 530 overlap audit](../../reverse-engineering/casino-road-overlap-triangle-audit-r314.md).
Neither result justifies moving a source instance or globally transposing its
matrix. The unresolved citywide question is correct retail visibility and
scene assembly, not a demonstrated coordinate error at these two locations.

The current host capture is in [review item 451](index.html#r403-current-host-drive-recheck).

## Full L4RA building-triangle / road-triangle scan · r464

The packaged viewer's `__nfsuFindPlacedPropRoadIntersections` diagnostic was
run at `(0, 0)` with a 10,000-unit radius, which covers the exported L4RA
stream's XY bounds. Restricting the scan to the `XB_` building namespace
produced 2,453 candidate placed groups; 589 groups had at least one pair of
building and driveable-road triangles that crossed in 3D at overlapping
height. The diagnostic records up to 100 crossing triangle pairs. This is a
broader geometric warning list than the prior route-corridor checks, not a
finding that 589 buildings are misplaced: it also flags shared boundaries,
foundations and structures intentionally built over or beside pavement.

I inspected a dense cluster around recovered route 716 at distance 424 in
host Chrome. The capture shows a close-frontage cobbled street without a whole
building across the visible driving lane. That one view cannot validate the
other 588 flagged placements. The route-corridor audit remains a separate
check of facade collision along the sampled traffic routes; it did not justify
moving any source building. The full-stream triangle scan and focused capture
are in [review item 493](index.html#r464-route716-placement-cluster).