Thanks for the explanation, personally, I am not sure the data could answer the question like:“hey after comparing 8 different model here, we are sure that model X is the generative process”. And building a 8*8 comparison table of all the models seems a bit difficult to reason. The mixed effect model is more like: “we assume every conditions are different, and build a full model the encodes all the information, then look at the posterior to see if they are really all different (i.e., no overlapping)” - basically you plot the posterior and see which ASCII-diagrams for different hypotheses it fits the best.
But since you have strong theoretical motivation, it certainly sound like a reasonable approach.