We currently have 4 raw SharedPromise<void> on ReshardingCoordinator plus 5 SharedPromise<ReshardingCoordinatorDocument> on ReshardingCoordinatorObserver, all with hand-rolled if (!isReady()) setError(...) blocks scattered across resharding_coordinator.inl and resharding_coordinator_observer.cpp. The same arguments that motivated ReshardingRecipientPromises (SERVER-123511) and ReshardingDonorPromises (SERVER-125598) apply here: idempotent fulfillment via the registry, single broadcast setError, recovery from persisted state therefore less chance of forgetting to fulfill a promise on a new path or new promise.

Introducing the same pattern will make the code more robust and similar to existing design patterns.