Why can Naive Bayes work well even when features are not independent?

A. Accurate class ranking may still result despite imperfect probability estimates B. The algorithm tests dependence explicitly C. It removes correlated features automatically D. Independence is never used