Great question. Discover doesn't build the questionnaire versions prior to fielding the study (as Lighthouse Studio does). Rather, Discover creates a unique questionnaire for each respondent as each respondent takes the survey. It uses a random number seed to start the algorithm, plus it also examines the responses to the Level Ratings questions (and supplemented by any attribute level preference orders you specify when building the questionnaire) to try to avoid entirely dominated concepts. That's a feature that Lighthouse Studio's CBC cannot do.
So, if you were to interview 800 respondents, you should see 800 unique versions of the questionnaire. Evidence to date through our testing suggests that Discover-CBC experimental designs are actually a tiny bit better on average than Lighthouse Studio designs.