You want to try out the long term effects of a different layout, and decide to serve 1 out of 500 users this new layout. You remember the randomly assigned 'beta' visitors, so that you can keep serving them the new layout and track their long term behavior.
If one of Google's bots is one of the 'beta' visitors, and another one is not, they might interpret this as if you're trying to mislead them?
What is the safest way from an SEO perspective to do this?
Google offers the testing that you are talking about -- [url]www.google.com/intl/en/websiteoptimizer[/url] -- so I doubt that there would be SEO implications unless the method you use to do the testing is severely out of the ordinary (not to be read as, "unless you don't use Google").