At Deathwatch we avoid calling a board dead after one failed request. We keep the checks public and use a measured failure threshold but I m curious what edge cases people think we should handle better: temporary Cloudflare blocks, expired domains, parked pages, or services that return 200 while being unusable?