Its unusual for FSR. But the upscaling tech roots in anti aliasing. DLSS was originally designed to be an anti aliasing tech, quite litterally in the name (Deep Learning Super Sampling), so it isnt out of the picture that FSR could be used for the same. It was only later marketed to get more performance at lower resolutions afterword. (Im aware the post is not your thoughts, just more of a response to the specific segment you pointed out)