If you are restoring Salesforce from SAFE C2C, there are a few things you need to consider.
As a rule of thumb, if the structure of the element or section of Salesforce has changed since the backup was taken, the restore will probably fail in part or in whole, the restore can only put data back where it came from. The backup will only know the data structure based on the structure at the time of that particular backup and cannot calculate changes in the structure of the live environment nor make decisions about where to try and put the data if the field has changed, moved or deleted. It is smart enough to detect when the data restoration failure rate may be too great and will simply skip these elements to save on API usage. This detection is currently at a top tier but is being developed to dig deeper into the objects and metadata to make better, more informed decisions.
Regarding Cases:
If a case is open in Salesforce then you can restore data / metadata into it. If the data structure of the case has not changed since the backup being used for the restore, then the restore should work fine. If the data structure of the case has changed since the backup was taken, the restore may still partially work but will still show errors due to the location of the destination changing or been removed. The functionality to review this before commencing the actual restoration is in progress, as is the ability to re-map changed or moved fields.
If a case is closed in Salesforce, you cannot restore into it. You can open the case in Salesforce and then restore data as per the above. You can then chose to close the case again.
If a case is deleted in Salesforce, you cannot restore it back to where it was or with the same identifier. You can restore a deleted open case but this would be restored back with a new identifier. For deleted, closed cases, you would need to export the data to a .CSV from the backup, manually change the status of the case from closed to open and then use the data import tool to get it back into Salesforce, again with a new identifier. The ability to restore the data is still dependent on the data structure being the same or similar as above.
Full Sandbox to Sandbox or full Production to Sandbox restorations are possible but are also subject to the data structure being the same or at least very similar or they will create lots of errors and/or fail.