SSI Web provides a design version based on the internal respondent number assigned to the person taking the survey. Online, these numbers are sequential, and versions are likewise given sequentially (respondent 1 sees version 1, etc.).
In the CAPI module respondent numbers are assigned at random, so version are also assigned at random. When a respondent number exceeds the number of versions, it's treated as if we just started over from the beginning, i.e. if I had 300 versions and respondent 301 takes the survey, they would receive version 1.