The most promising upscaling technique is the neuron network - the one in that google link, but there are also others; result are astonishing, even if the final result is not really an upscale, but more a reconstruction; but, who cares, if the result is great? Let’s wait for someone who would bring these kind of techniques in AviSynth - it will happen, sooner or later…
Meanwhile, we should rely on the old techniques; I think mixing different ones could lead to a better result than using a single one. I developed one on my own, simple, but effective - the PaNup: https://forum.fanres.com/thread-25.html - I used in many projects, and it works quite well; can’t say how much can improve the quality in comparison of a single upscale, but there is an improvement; I encourage everyone to take a look and try it - it’s easy, and doesn’t require a specific tool like AviSynth…