Finding the proper scope of a file collapse operation

If the referenced frames are keyframes, it might be easier/faster to extract the frames than seek to the keyframe, decode N frames and then start extracting frames. (Just speculation.)