BfA WMO backport to Wrath 3.3.5a: geometry disappears at certain camera angles what visibility data are we missing?

[NordReborn]

I’m developing Gaia, a tool that backports modern WoW world data into the Wrath 3.3.5a client.

We have successfully converted Kul Tiras/Boralus terrain, placements and a large amount of its architecture. The city loads and is explorable, but some sections of WMO geometry disappear and reappear as the camera moves. This is not an FPS hitch: it is repeatable, viewpoint-dependent geometry popping.

The same failure occurs:

• On a converted Kul Tiras map with surrounding ADTs present.
• When the Boralus tile is installed on a separate test map.
• At a distance and close to the affected structures.
• With the player stationary while rotating the camera.
• After removing obvious misplaced modular wall objects.
• After removing the modern volumetric-fog M2.
• After rebuilding ADT MCRF WMO coverage from final MODF world-space bounds.

What we have already checked:

• Final MODF placement bounds.
• MCNK/MCRF coverage for every intersected terrain chunk.
• WMO root and group bounds.
• MOBA batch unknown_box values calculated from the final MOVT vertices.
• Modern placement scaling, including scale-baked WMO variants.
• Terrain holes and cross-ADT placement context.
• The problem remains on the fuller Kul Tiras map, so an isolated ADT lacking its neighbours does not appear to be the cause.
• Materials are imperfect, but the affected geometry is genuinely being culled rather than merely displaying the wrong texture.

A native Wrath Lion’s Pride Inn originally showed similar camera clipping when inserted through the old Gaia path. Reconstructing its placement/spatial data fixed it completely. Boralus, however, still has residual popping after applying those rules.

My current suspects are:

• MOGI/MOGP group flags or bounds.
• MOPV/MOPT/MOPR portal geometry or relationships.
• Indoor/outdoor group classification.
• Modern WMO group/LOD selection semantics that Wrath does not understand.
• Converted WMOs containing groups that should be flattened, remapped or selected differently for 3.3.5a.
• Some additional Wrath-era visibility structure that must be reconstructed during retroporting.

For people experienced with Wrath WMO internals:

After root/group/MOBA bounds and ADT references have been rebuilt correctly, which WMO fields can still cause individual groups or batches to disappear at particular camera angles?

In particular:

1. Does Wrath perform portal-based culling on outdoor WMO groups, and can invalid modern portal records cause this behaviour?
2. Which MOGP/MOGI flags materially affect camera or group visibility in 3.3.5a?
3. Are modern WMO LOD groups represented in a way that must be collapsed or removed when targeting Wrath?
4. Are there known differences in portal-plane coordinate conventions, group indices or relation directions between BfA and 3.3.5a?
5. Is there a proven tool or audit procedure for determining exactly why the Wrath client culled a WMO group?

I’m not looking for a speculative “increase draw distance” fix. I’m trying to identify the precise structure that Wrath is consulting when it makes the incorrect culling decision.

Any format documentation, old repair scripts, known client behaviour, or small test WMOs demonstrating the rule would be extremely useful.

Cheers all