Why do many real-world type checkers use bidirectional type checking instead of pure inference or pure checking?

A Because it is the only approach capable of type-checking dynamically typed languages B Because it guarantees that type inference always terminates in constant time C Because it eliminates the need for a symbol table during compilation D Because it combines a synthesis mode (compute a type from an expression) with a checking mode (verify against an expected type), handling polymorphism and literals with fewer annotations