Currently, the analyzeShardKey command doesn't assign an _id to the split point documents that it generates. As a result, if there is a failover during the inserts, some documents can get inserted multiple times due to the auto retry. To solve this, each document must be assigned an _id and DuplicateKey errors should be ignored just like what that's done in the QueryAnalysisWriter.