It's a part of an initiative of bringing wtperf YCSB closer to MongoDB YCSB (WT-17498). 

By default, MongoDB uses a scrambled Zipfian distribution for YCSB, while wtperf currently uses a regular Pareto distribution, which increases the load on the first key (1 out of 20 operations still targets key #1). The scrambled implementation additionally hashes the generated key, so the hottest key ends up somewhere in the middle of the range instead. While implementing a Zipfian distribution requires more effort, it is relatively easier to simply make the Pareto distribution scrambled.