in the Benchmark article http://drivendata.co/blog/rinse-over-run-benchmark/ ,
there are phases randomly dropped after dropping the final_rinse phase.
If there is only 1 other phase remaining in a process, why would I drop a phase and therefore the whole process? I would make a restriction that only in processes with more than 1 phase, the random dropping of a phase is performed.
In the Benchmark this has not been accounted for.