Apart from the obvious possibility of IP ban (that there's only one way to circumvent), I think parsing is the most brittle part of this. If I needed to make so

Apart from the obvious possibility of IP ban (that there's only one way to circumvent), I think parsing is the most brittle part of this. If I needed to make something that would just survive without me rewriting it all the time, I would do away with parsing somehow, at all. E.g. instead of trying to parse anything, just load it in a browser (headless, or even real) and just somehow see where the images are (from the requests log, or from the screenshot).