No, in a Randomised Control Trial, you randomise into arms on the basis of intention to treat. So, for previous CS you have two groups: one having repeat elective CS, the other attempting VBAC. Then you look at the outcomes (maternal and fetal) of both groups, regardless of whether they were successful in achieving VBAC. This is the best simulation of what happens in 'real life', since not every woman who embarks on VBAC will get one (some will have an in-labour CS). The maternal benefit of attempted VBAC over repeat elective CS remains, even after accounting for the % of women who will attempt VBAC but end up having CS. (And some of the risks associated with previous CS are not mitigated by planned elective CS, since planned elective CS is not completely protective against rupture).

Similarly, with comparison of breech vaginal delivery, you have to compare planned CS for breech, and planned vaginal delivery for breech, then look at outcomes (maternal and fetal) taking into account that not all plans (CS or vaginal) will, in real life, go according to plan. Thus some *planning* a CS doesn't mean you will get one, nor does *planning* a vaginal delivery mean you will get one. Since no one can guarantee CS or vaginal, the only meaningful comparison is the intention. Now, if you have a big cross-over then you have to account for that. Maybe the threshold for in-labour CS was too low (or too high) in the vaginal breech group? Maybe that effects the outcomes?