A team achieves high line coverage but important failure behavior still breaks. What should it conclude?

1. Tests should be removed once the percentage is high
2. More coverage always eliminates flaky dependencies
3. High coverage proves every business requirement is correct
4. Coverage shows execution reach, not that assertions verify all required behavior