Under certain circumstances, a node that is currently acting as Primary - Host A - dies suddenly during an index build process resulting in an election of a new primary - Host C. The index build then will conclude on the node that was previously primary (Host A). As Host A is not the Primary the index build isn't replicated to the other nodes and resulting inconsistent indexes between the cluster members (Host A versus B and C). Additionally, this behavior is detailed in SERVER-14746.

In these cases, it might be useful to have a command to view inconsistent indexes or indexes across the nodes in a replica set in order troubleshoot differences in behavior by running rs.checkIndexes(), for example. Such a command might also be able to be leveraged on our Cloud offerings in order to proactively identify such circumstances before an election causes a change in behavior for the customer's application.