great article, the problem I found in those cases is that some pages like immobilienscout24 sometimes take a lot to load, and that increases the cost. Also in my case I found some websites (not immobilienscout24) blocking my requests, so I ended up implementing proxycrawl to unblock the requests, that allowed to bypass the blocks of google cloud but still I faced the loading times problem of some of the websites

Neat article. My only gripe is that the cost is a function of how much content you scrape. If 3 minutes turned into 30 minutes it would cost more, regardless of the tech stack. Also, since web scraping is IO bound the only thing that really matters is how fast GCP compute servers can download the content.