> I’m skeptical that the problem they are trying to solve is truly unreasonable bandwidth demands.
Not necessarily bandwidth demands so much as processing demands. Scrapers have a tendency to hammer on parts of web sites that are computationally expensive to generate - e.g. search results, diffs and blame views in git forges, sorted/filtered/paginated lists, etc. Ordinary users may click a few of those links for things they want to see; scrapers will try to request all of them, even when 99% of them are redundant.
What’s more, they will scale up with increased resources on the site.
If you redline at 20 searches a sec, and put in 4 more workers, suddenly you’re serving 100r/sec to the bots, paying 5x for it, and your users are still seeing shit qos. I've seen multiple cores of nginx saturated just dealing with one dos/crawl run on a somewhat high profile site.
Not necessarily bandwidth demands so much as processing demands. Scrapers have a tendency to hammer on parts of web sites that are computationally expensive to generate - e.g. search results, diffs and blame views in git forges, sorted/filtered/paginated lists, etc. Ordinary users may click a few of those links for things they want to see; scrapers will try to request all of them, even when 99% of them are redundant.