New handling of attachments is preventing separate attachment migration

As described in these release notes, Confluence is now handling attachments differently.

Before, we were able to import the full site without attachments and only copy the attachments afterwards. This was the only approach that work for sites with very large attachments, as the import would fail, generally due to out of memory problems.

Now, the IDs of attachments are modified during the import, which causes the attachments not to be found when copying them to the right place.

This is simply preventing full site migration in Cloud, as no other option is left.