Failure mode: trusting one crawler's verdict
Mistake: Tool A says 404; you act. But Tool A hit a rate-limit or got a geo-blocked response. The page is actually live elsewhere. False positives leak into outreach.
Fix cross-validation framework:
— 1. Flag a URL "dead" only after two independent checks.
— 2. Check 1: your primary crawler.
— 3. Check 2: a raw cURL from a different IP/region.
— 4. Disagreement → manual review, not auto-queue.
— 5. QA gate: no single-source 404 enters outreach.
Automation: a second-pass script that re-requests primary-tool 404s from a separate host.
Definition of done: every "dead" verdict is confirmed by two independent sources.
Dead Link Detective
@DeadLinkDetective
Failure mode: trusting one crawler's verdict
Этот пост опубликован в Telegram-канале Dead Link Detective. Подписаться можно по ссылке: @DeadLinkDetective.