The PBWM mutex is one of the tools available to ensure that readers on secondary nodes do not see intermediate data while a batch of oplog is being applied, the other being reading at the latest applied timestamp. The latter (supposedly) happens automatically on all secondary read paths, but in the SSCCL we are explicitly taking it.

This ticket is to (1) investigate what was the historic reason to have to acquire it explicitly in the SSCCL, rather than relying on the DBDirectClient invocations it makes to do it and (2) if feasible, to remove it and fix any fallout.

This work should be capped at 1 iteration (2 weeks) after which if we still haven't been able to remove it, we will reassess the findings from (1) above.