Measure twice, fabricate once. Or twice?

I spent so much time this weekend working on the PCB layout I decided to do something a little different this evening: I printed the board layout on paper in a 1:1 scale and poked holes through the paper to see if they lined up with the holes in the original calculator PCB. Bearing in mind the last time I mis-measured before fabricating a PCB, I'd measured everything quite carefully. I was feeling quite pleased with myself until one of the holes marked on the paper didn't have a hole in the PCB beneath it. It looks like I got the horizontal position correct, but somehow I got the vertical position of that hole off by 2 mm. Oops.

Now I'm carefully cutting out holes where each of the keypad patterns should be and peeking underneath to see if I got those lined up. Thus far they do, but rather than spot-checking a random sample I think I'm going to check each and every one.

Maybe I should make the keep-out areas around the holes a bit larger. That way, if I get one or two off by a little bit I can ream out the hole and not risk tearing up tracks or the interior ground plane in the process.

I'm half-tempted to have a 2-layer keypad test and physical dimension check board fabricated once I finish checking against the paper prototype. I'd take the board I have, drop the two inner layers and rip up all the tracks, leaving only the component footprints. Then I'd reconnect the keypad in a simple grid pattern and bring those circuits to the upper edge where I could reach them with an ohmmeter. But if I get things right the first time it would be a waste of time and money. What are the chances of that happening?

Tempting, but probably not worth the cost. With a HASL finish it'd cost about $32 and take a week including delivery, rather than $90 and 10 days for the 6 layer board and stencil. But HASL -- Hot Air Solder Leveling -- isn't flat like the Electroless Nickel Immersion Gold finish I'm planning to use and thus would invalidate any keypad testing I did. An ENIG finish would raise the cost to about $50, which is getting pricey for a throw-away. I think I'd rather bet on getting the board close enough the first time.