> I remember reading somewhere that Google doesn't promise not to visit pages it's excluded from; it only promises not to put them in the SERPs.
I hope you didn't read that here, suidas!
The opposite is the case. If Google sees a link to a /robots.txt excluded URL, then Google will not fetch it. It can still list the URL in the results without having to fetch the page.
If a URL is not /robots.txt excluded, but has a META robots tag with 'noindex', then if Google fetches the URL it will not be listed.
> sure this has been tried, and Google's caught it
That someone has a Web site, and that there's some part of the Web that shouldn't be crawled? I don't see how Google would see a quality issue there, unless they didn't like the site they were crawling.
The 'dual site' approach using Robots Exclusion Protocol would sacrifice links though.