What's the most dangerous silent error in PDF-to-Excel work?

I keep seeing spreadsheets that open cleanly but still contain missing rows, shifted data types, duplicates, or totals that no longer reconcile. I'm building a repeatable QA workflow around those failure modes. For people who review extracted tables: which check catches the most expensive errors in your work, and which one is still annoyingly manual? I'll use the answers to prioritize the next update to PDF to Excel QA Toolkit.