In Jenkins it is possible to assign labels to nodes, e.g. somenode to nodes and then one could call somenode in the pipeline and then Jenkins will run the build on some of the nodes.
The assumption was that Jenkins would apply round robin, i.e. the number of builds that is run on node 1 2 3 4 and 5 is equal, but actually the majority, I.e. more than 90 percent is run on node four.