To address a high phylog:oplog ratio, a mitigation was put in place in SERVER-114491 to reduce the memory_page_max setting for the oplog from 10M to 512K.

Per Alexander Gorrod in SERVER-120876:

We put a workaround in place for that in SERVER-114491. Ideally, we'd be able to revert that change once this mechanism is in place.

As part of SERVER-120876, we're adding a mechanism to control checkpoint frequency based on oplog volume, but the memory_page_max setting remains unchanged.

This ticket tracks revisiting the memory_page_max setting — either reverting it to 10M or setting it to a more appropriate value based on a few perf runs.