For IGH default seed changed (from #####-##### to ######-######) which changes
segmentation results (and improves them!).

Should we provide the seed to be used as a parameter to avoid such changes? Or
should we let the test as it is to monitor changes in default behaviour?