Anisotropy and XKS-splitting from geodynamic models of double subduction: Testing the limits of interpretation