Are you pooling the data from the 50 different participants? Because you are already working with a hierarchical model I had assumed that you using a partial pooling over participants, but I don’t see any participant-level parameters. If you are pooling, that complicates things further, because now you’re not working with a standard right-skewed RT distribution, but the sum of 50 such distributions. Without knowing how participants vary (e.g., the variability of individuals’ mean RTs), I wouldn’t know what to expect of such a distribution or how to go about modeling it (without acknowledging the existence of the individuals).
So let’s ignore the pooling issue. If I have data from a 2x3 design but can’t successfully characterize the data from 1 of those 6 conditions, then I’m not going to move on to the full model, because it’s not going to get any easier. That’s all I meant.