Finally, I discovered that my admins are facing the issue/limitation of V9 discussed here: veeam-backup-replication-f2/vbr-console-restore-proxy-t32565.html
Which help me understand the lack of informations in a multiple restore sessions logs I had.
The long story of 2 discovered limitations:
The helpdesk tried multiple times to restore files in a remote office (Remote Proxy/Repository) using the central Veeam admin console located in HQ where all the backup jobs are configured. They launches the restore wizard using the console from the HQ, the FLR component is mounting the backup to the mount server which is the machine running the veeam console in HQ. Then the FLR stop because of timeout (more than 30 min because of the slow WAN link). Because the logging for a restore sessions is not verbose and the log level is not configurable, there are not enough detail to help me tracing what they did exactly and understanding that there was an issue. Some of these restore logs in the history are reporting a successful restore session, which was not the case, and other are reporting an inactivity timeout.
Unfortunately, the best solution I found is to install a remote standalone backup server that import the backup files from his local repository so the backup admins and the helpdesk can restore files from this remote office without lags and timeout (Note: I also tested to install only the Admin Console on a computer in this remote office but the Restore Wizard was running very slowly). I'll use the same solution for my other ROBOs with a slow WAN link until an update or a new release of Veeam B&R allow us to specify the mount server for a restore job. I thought that the improvements announced in v9 regarding ROBO were more complete.