Chord Dave Replacement

"I’m not sure if there are algorithms that can add high-frequency information that wasn’t captured in the original sampled result"

This is the most important statement , I believe upscaling or upsampling whatever you choose to call it is not more than a voodoo.
You can't add information that is not exist at the original recording it's not more than a manipulation process that might degrade the sound. 

That statement makes sense in an analog paradigm. But digital, potentially, can change things.

Think of video. Upscaling an image (well) can most certainly improve the quality. That’s not to say that algorithms can restore *the same* detail that was originally present but lost in the lower-resolution digital capture. However, it can appear to be more like the original to our perception.

This is because scaling is now more sophisticated that merely "averaging new dots between the old" (what used to happen). Sony has had AI tools analyze millions of images in high resolution, and compare with the low-resolution rendering. Based on patterns, the AI learns to "restore" back detail in the low resolution images that appears to our senses to be real detail. It technically isn’t 100% accurate to the detail that was there in real-life, but it’s "more accurate", or at least it appears to be so to our vision, than the low resolution signal. And what’s especially impressive is that they can do all of this on an affordable chip, and in real-time, in their higher-end TVs and projectors.

Upscaling (and potentially upsampling) audio signals is exactly the same. There’s no reason why good AI couldn’t be trained with pattern recognition over time to perceptively restore natural detail into low-resolution digital audio the same way.