It can be applied empirically to speed up large number computations even in the absence of full theoretical proof.
Only two problems:
1. You would have no confidence in the results.2. The point of the paper is advance knowledge, not guesswork.