Variable identityTransformsConst

identityTransforms: inspect.sceneModel.Inspection = ...

Flags every SceneTransform whose matrix is the 4×4 identity (within float epsilon). Identity transforms contribute nothing to the world transform — they're either an artefact of a loader that always emits a transform per node, or stale state left behind after pose decomposition.

The matching dropIdentityTransform re-parents every referencing mesh / child transform to the identity transform's grandparent (preserving world poses, since the missing factor is identity) and then destroys it.

Purely advisory — leaving identity transforms in place doesn't break anything visually, just costs a few bytes and a matrix-mul per frame.