Your solutions depend on what version of SSI Web you are on, but I'd always strive to send everyone through one survey just so you have unique respondent numbers. If you split people out you'll have four respondent 1's, four respondent 2's, etc.
On earlier versions of the software, you could just use some skip logic to move people around, i.e. if you chose the first radio button of Q14, you should see the first introductory text/filler page and then continue on through the first MaxDiff exercise. If you chose the second radio button, I'll skip you over the first exercise to a second text/filler introduction and you take the second MaxDiff exercise.
In later versions of the survey, you could put MaxDiff exercises into randomized blocks and control what block a person goes
From a design perspective, I don't think you'll see much practical effect if you have "holes" in your designs, i.e. nobody saw version 287 of the third MaxDiff exercise. Each version is constructed to be balanced and orthogonal.
