Two years ago I did a back-of-the-envelope calculation suggesting that if a well-run state is to use surveillance to prevent or investigate politically motivated violence, then it has to spend considerably more than €3000 per year and person to get results (where
person means anyone who might be willing to help plan or carry out political violence).
The limit is higher: The €3000 did not include the price of xkeyscore and the other ineffective programs that have since come to light. Most of that cost is borne by the USA, but the
500 million records supplied monthly by Germans will have cost something.
I suppose it's theoretically possible that the names of known Nazis weren't added to the list of suspects, but I can't bring myself to believe it. These are people who gather in the woods to practise with guns, and their names are in a database run by someone who supplied data to the NSA. They must have been added.
So xkeyscore had the names of Mundlos, Böhnhard, Zschäpe and probably everyone who helped them, and it found nothing, neither before any of the killings nor afterwards.
It found one thing, though: Laura Poitras had 400 out of 400 points on the risk scale, and in the event she has turned out to be a great risk to job security at the NSA.