OpObservers independently recompute and lookup global state. Like inRecoveryMode which requires several calls to ReplicationCoordinator. The onUpdate chain computes this state at least 4 times to filter events.

Currently the OpStateAccumulators are implemented post-7.0 (SERVER-76369) in a handful of methods (broadly speaking, the CRUD methods) that functionally require downstream OpObservers that observe the opTime allocated by OpObserverImpl.

This ticket would extend the accumulators to be used for performance improvements on top of its existing purpose of functionally breaking dependencies.