Record the pose-composition fix and the remaining Godot skinning-import blocker

TODO.md

@@ -775,11 +775,16 @@ blocked rig can no longer block the whole release.
       (pose-rotate → apply modifier → reset pose → edit-mode rotate arms → re-add modifier →
       export `export_apply=False`) now yields a correct arm-down base: Blender deformed x ±0.221
       and the rest bone tail points down. Still local (`tmp/`), not shipped.
-- [ ] Re-derive the pose composition for a non-180° rest. Godot's `set_bone_pose_rotation` is
+- [x] Re-derive the pose composition for a non-180° rest. Godot's `set_bone_pose_rotation` is
       rest-relative, so `rest * Quaternion(axis, angle)` only reproduces the old rig by accident
-      (the old rest is 180°, so `rest² = identity`); the Quaternius ~76° rest breaks it. Derive the
-      correct composition (candidate: `rest.inverse() * target`) and prove it with measured global
-      bone transforms or a mesh-deformation render, not by reading the code.
+      (the old rest is 180°, so `rest² = identity`); the Quaternius ~76° rest breaks it. Fixed in
+      commit `8b1bf73`: `rest * Quaternion(rest.inverse() * axis, angle)` — backward-compatible
+      (the old 180° rest maps the axis to itself). Godot 277/277.
+- [ ] Godot skinning import: the arm-down GLB is now provably correct (raw vertices x ±0.175,
+      rest tail down, inverse bind matrices arm-down; Blender renders arm-down), but Godot 4.7.2
+      still renders the mesh in a T-pose. This is a Godot glTF importer skinning issue, not a
+      Blender bake or pose-formula issue. Next: isolate with a minimal 2-bone glTF, check the
+      Godot issue tracker, and try alternate export settings.
 - [ ] Add regression tests: a bone with a ~76° rest reaches a known global orientation; the
       Quaternius joint map resolves; the 13-core fallback poses exactly as it does today.
 - [ ] Verify Idle and Walk on the 62-joint rig: loop seam, foot contact, root motion and frame