SERVER-40250 showed the high contention on the same mutex under heavy concurrent w:majority writes. The user reported 7% CPU usage with 256 threads inserting small documents on a 16-core machine. The bottleneck was _wakeReadyWaiters_inlock which implements a naive O(N) algorithm to examine waiters, where N is the total number of all pending and ready waiters. 

We need to verify the bottleneck in a 2-node or 3-node replica set and add Genny test to evergreen.