When plotting the benchmark (and other domains) with the rimwidth set to 3 for example, the depths of the plot look very odd. I think there is a bug in the ordering of bdy points in nemo_bdy_oder() which doesn't seem to be an issue for the outer rim.