When heartbeat response has a newer config, the node tries to install the config immediately without updating the heartbeat state. SERVER-47949 doesn't allow a new config during catchup mode, but catchup mode relies on the latest heartbeat state. Thus the primary will never exit catchup mode if it sees a newer config in heartbeat responses.

We should update the heartbeat state even if topology coordinator sees a newer config.

It's less ideal to move the logic of SERVER-47949 to topology coordinator where it decides to install the new config because some of the information is only available in replication coordinator, like whether the node can accept writes.