Issue Summary

This is reproducible on Conf Cloud: Yes

Take backup from Cloud to Server and then back to cloud is a use case from https://confluence.atlassian.com/cloudkb/migrating-from-confluence-cloud-to-confluence-cloud-691011969.html  

Steps to Reproduce

1. Take a backup of Confluence Cloud for Confluence Server. 
2. Import on Confluence Server and run a migration to Destination cloud via CCMA
3. You will end up with duplicate attachments because of the file store IDs in Cloud backup for server. 

Expected Results

No duplicate attachments should be seen. Cloud backup created for server should not have these IDs to begin with.

Actual Results

The duplication happens because the data in cloud has filestore_ids. This gets copied onto contentproperties in server.

Now, when we migrate from server, we migrate these entries and also create new entries for every attachment with a new mediaId that gets generated during attachment upload.

Thus resulting in contentproperties in cloud having two entries for every attachment contentid.

Workaround

Use the dark feature

migration-assistant.enable.original-xml-export

 on intermediate instance.