Agree with your point though, and it may just be one of those impossible questions which is why the patent office struggles with how to make these decisions.
But when it comes to money-making patents, I think of technological development as an optimization process where everyone has the same objectives and a pretty limited search area at each point in time (available technologies you can use, textbook knowledge you can draw on). I'd generally expect any smart and dedicated person working on the problem to find the next best next steps sooner or later (certainly far less than 20 years).
Maybe take a 3rd party who is unaware of the patent/novel idea and then asking them how to solve for some generalized version of the problem that a patent states it is solving and see what happens.
Similar to black boxing that companies will do with tech that may have been shared under restricted terms.
Just do peer review and get experts' opinions. Don't tell them how to make it work, just tell them what it does. E.g. make the inventor provide a carefuully-worded abstract. By the way patents that claim the category of problem itself as the invention (as opposed to the specific method for solving the problem) are another big problem that needs to be eliminated. Those broad first claims are the ones shot down in challenges. For example "use a computer to processs transactions", or even "use a convolutional network to classify faces". These are not inventions; they are problems that still need to be solved, and it can be done many ways.
Anyway if you're uncertain about 6-months, how can you justify giving them 20 years of monopoly?