Richard,
While I am not a mind reader, I think that JS comments about money were directed at the publisher of the DVD, and not you.
I agree with JS that the DVD has many errors and I should have also pointed out the limitations of the method. The smaller the number of teeth the gear has the more error there is in that method. For some applications it will produce gears that are usable, although they may not be technically correct .
Getting back to Dreeves original question. If you are making gears for a non critical application, then it is possible to use a cutter that is incorrect for one of the gears. By non critical, I mean that you can tolerate some extra noise, that the gears will not have a large load on them and that you can tolerate some extra backlash. You may also have to make some adjustments in your center to center distance on the gears to get them to run smoothly.
Using your example of a number 3 cutter for a 40 tooth gear and a number 6 cutter for a 20 tooth gear.
If both gears are cut with a #3 cutter, the gear teeth will have an interference with each other and you would have to increase the center to center spacing of the gears or the gears will bind.
If both gears are cut with a#6 cutter, there will be additional backlash in the gear train, but they will mesh at the normal center to center gear spacing.
In both cases they will make additional noise, but this will probably not be a problem at low speeds. If loaded heavy there will be additional wear due to sliding contact on the gear teeth. Wear may not be a problem is you are building an engine that will only see a few dozen hours of operation in it's lifetime.
I a choice is to be had, I would probably pick the higher number cutter, then you could at a later date cut a new gear to specification and directly replace it.
So, it depends on the application. I would not have a problem doing for something like Chucks single that you built, but I would not do on a high speed IC engine valve gear train.
Gail in NM,USA