A number of issues identified by implementing shard split stepdown passthroughs:

• Waiting for recipient nodes to reach block timestamp doesn't work because no recipient nodes are listed in local members after applying the split config
• Similarly, we attempt to rebuild a recipient connection string after a split config has been installed, which is impossible since the recipient nodes have been removed from the local member list
• Installing a split config on top of a pre-existing split config may result in strange replication errors ("installed the config, but unable to get majority commitment")