It's only possible if you can identify large portions of the 50k original lines as having been previously implemented by other components (python modules, microservices, etc.), or that large portions are dealing with cases that are guaranteed to no longer arise (so you either produce different results or error out if you detect them).