Am I correct in reading that this vector only allow the attacker to determine the private key of an address that has been used to sign a transaction? IE, if you use all the inputs of an address in the transaction and not reusing any addresses even a compromised ECDSA module would only net the attacker your now empty address.
Well, the paper isn't really published yet, but as far as I can tell, this seems to be the case.
Honestly, the whole issue is interesting, but not much more.
All it really shows is that you can actually use the transaction signing part of cold storage to get information out of an otherwise sealed system.
Then again, that's more or less Captain Obvious speaking