The typical distance constraints for attaching the cloth haven't been quite right, leaving it with an obvious tendency to rotate or spin around the curve.
The problem I've broadly solved with additional geometry. Branches from the main curve maintain their orientation while giving the cloth something to attach to. This works, but it strikes me that there are likely less convoluted methods of achieving the same result.
I've attached a clip of a spinning constraint, versus not spinning branches.
How wrong am I?
Cheers,
Edit
I should qualify that the “Yes” label in the video implies yes that you're looking at the “good one”, not the one with the spinning constraints.
