Video Frame Interpolation (VFI) is an important technique for temporal video enhancement, enabling applications like slow-motion effects, efficient animation rendering, and broadcast frame-rate conversion. Driven by advances in optical flow estimation and deep neural networks (including flow-based, kernel-based, and transformer architectures), modern VFI methods can predict complex motion and handle occlusions. However, evaluating the perceptual quality of interpolated content remains challenging. In our previous work in archived motion picture restoration we used vector field divergence to detect spatial irregularities in optical flow fields that indicate temporal inconsistencies. We find that using this feature to weight PSNR in a principled manner, we can identify and emphasise problematic motion patterns – such as unnatural divergence or convergence – that most degrade perceptual quality in interpolated frames.