Deleting the data directory on master and reindexing causes slaves to be out of sync (if replication doesn't take place before a master instance restart)

This issue occurs only when the data directory is deleted manually on the master instance.

Here are the steps that I used to replicate the issue:

Master - Shutdown
Master - Delete Data Directory
Slave - Replication triggers, but no connection to master, replicaiton fails
Master - Start Up
Master - Index all docs
Master - Shutdown
Master - Start Up

Expected: Slave - Replication triggers, connected to master, new index replications, master & slave in sync

Actual: Slave - Replication triggers, connected to master, error seen in the logs: "SEVERE: No files to download for index generation: 2"

• Nothing will replicate even if master generation > slave generation

Workaround:

• Must trigger another commit on master after the restart
• Replicates even if master generation < slave generation (and gets the generations in sync)
• Replicates when master generation > slave generation