The "Higher-Order Test Generation" paper questions the practicality of sound concretization and higher-order test generation for large applications, suggesting instead that testers provide manually the names of functions to be concretized. But we are more ambitious, and ask that you describe in this OP an approach to enable complete automation in higher-order test generation. How would you do this? Can you automatically identify functions to be concretized in a real system? Can you find an efficient way to capture all input-output value pairs of functions/instructions? Provide sufficient detail to convince the reader that your approach will work in practice, but no more.