I was looking at the latest issues/commits and came across ITS#8521 which discusses the correct sequence of setting up the serverID, the syncprov overlay and uploading the actual data in a replication scheme. So there, the conclusion is:
"Bottom line, *always* inject the serverID and syncprov *before injecting
some data, or be ready to do an update after having injected syncprov
and serverID"
This made me wonder how can we accomplish this with the replication of cn=config, and if the replication of cn=config requires any extra precaution at the initialization phase.
Nikos