Interesting. I see the library they are relying on, recipe-scraper, uses the exact approach we mentioned (ld+json + custom scrapers as fallback) so the pros and cons discussed apply. Good to see work in this area!