In the process of developing the new shard versioning protocol we overall agreed that waiting for the configTime would leave us in a position no worse than today where the secondary node has to issue a refresh on the primary and then wait for replication.

However, we didn't fully explore all potential options and thus we can't tell if the following statements are fully true:

• Will convergence of the secondary with the primary's state not be slower than today in regards to the shard version
• Will we not be worsening overall the amount of time spent waiting
• Is there any other scenario we haven't thought of that could lead to a slow convergence