You described my contributions as "odd theories" from the refinements thread, while claiming those points were "actively discussed."
They weren't.
I was the main (often only) person in that thread consistently posting detailed tests: frustum roll from aircraft pitch due to camera mounting, 1:1 bank derotation calculations, stitched cloud motion analysis, elevation tables, and the Go Fast control test from the same flight/same system.
You yourself identified the slight clockwise rotation issue in the first 22 seconds as one of the two main problems needing refinement. That's not an "odd theory" — that's something you highlighted and attempted to fix with tunable extra rotation.
You also replicated the exact "katana shape" with your feature tracking, which further supports the frustum/pitch effect.
If we're going to discuss facts, let's discuss the actual data: the lack of demonstrated need for the extra rotation mechanism, the Go Fast test failure, and why triangulation failed in your recreation.