In the CHSH inequality, we construct an experiment whereby two observers each receive a particle and measure two given properties of their particles, for which the outcomes are ±1. We then consider the expectation value of some quantity defined by:
$\langle S \rangle = \langle AB - Ab + aB + ab \rangle$,
where A and a correspond to the measurement results for the two properties measured by the first observer, B and b for the second.
My confusion stems from the fact that each observer only measures one property at a time. Thus, it follows that the values for A,a/B,b must be the values that each observer would have gotten if they had measured the corresponding property. This will be equivalent to the value that each observer actually does measure if we assume realism, which is valid as this is one of the two assumptions we make for the CHSH inequality (the other being locality). However, when this experiment is actually carried out in the lab, each observer only measures one property, thus, for each iteration, there only exist 2 terms that can be plugged into S. Is this a problem and if so how is it reconciled?
EDIT: Resolved in comments below but idk how to resolve a post without a formal answer.