How a small daily crawler pointed at New Zealand's insurance auctions quietly grew into a three-year, 660,000-record public dataset. Scraping, GitHub Actions, cleaning, and the gotchas along the way.
