In my thinking, Corvus is right on the money. There's no point in discussing what a person should do when a dog attacks, because I don't believe Roscoe is acting in a "normal" way, even including the spectrum of "aggression."
To lay my cards on the table: I specialize in aggressive rehab. I take the dogs that other rescuers won't touch because of known aggressive behaviors. I also take dogs from owners, some of whom are rather advanced dog trainers such as goosedog trainers and stockdog trainers.
The first thing I do when deciding whether to take a dog on, is carry on a dialog with people who have experience with the dog's behavior. My experience allows me to pick up signals from the things people say, and from how the dog is described to behave in certain situations, that help me evaluate whether I can simply advise, or whether the dog should be removed from the situation, or whether it's a situation that I do not want to get involved in (in which case, I will advise getting the help of a certified veterinary behaviorist).
Next I meet the dog. I meet with more aggressive behaviors at this point than at any other time in a training cycle - obviously because I don't know the dog, but most often because the dog's behavior is linked to his or her relationship with the owner. If that's awry, the presence of a stranger adds fuel to the fire.
I don't look for "attacks." If you want to ask, "How many times has a dog attempted to bite you?" I'd say, maybe half a dozen or so. I've handled about two dozen specifically aggressive dogs in the course of my work, and the half a dozen bite attempts actually represent, in a couple cases, multiples from the same dog.
But part of doing what I do, safely, means being very alert for signs that a dog is past his or her comfort level and knowing how to defuse that tension. Very rarely does it involve any kind of pressure. If I can get it very, very early, I can ask for the dog's attention and put a little verbal pressure on the dog. A herding trainer friend of mine says, "Ask, demand, withdraw."
That means if the dog is stressing I give him a chance to think it through - get the brain going with a request I know he can handle, a mild correction (this does work) and repeat the request to see whether he's got anything to give, and then take whatever the dog can give, even if it's not the exact thing I asked for. Wow, that's really hard to verbalize. If none of that works, we're outta there. We take five, or whatever it takes for the dog to regain the ability to think.
Suffice it to say that the safest thing for anyone but a pro, is to withdraw immediately. Don't put an aggressive dog into situations it doesn't have the tools to handle.
That's an important thing that a lot of people can't seem to realize about aggression. It's not disobedience, like a child throwing a temper tantrum. Aggression represents a dog that doesn't have the ability to deal with whatever is going on at that moment, in a socially acceptable way.
Rehabbing aggression, when I do it, means trying to find the tools that the dog needs, and then developing them and reinforcing them. By the time a dog leaves here, they typically don't just have the ability to survive the situations that set them off before - they enjoy those situations.
So back to Roscoe. The problem here is that I am not gathering from the information given, any kind of pattern. The biting is never what I look at in these cases. When a dog bites or even "attacks" tells me nothing, really. What I need to know is where the adrenaline is ramping up, where the comfort level is being crossed, where the dog's brains fall out his ears and goes into "fight or flight" or whatever other low-level brain process drives his behavior.
Poor Roscoe is either not giving any clear signals prior to his going into the zone, or his owner is missing them. I don't mean hard stares and growling - that's a dog already in the zone. Given the severity of the incidents, I would not think the owner is missing signs of stress. If it's that there truly isn't any buildup, this is a combination I don't like to see if there are no medical issues.
Idiopathic aggression is untreatable by amateurs. I even consider myself an amateur in this respect. I'd strongly recommend the owner to get a consultation with someone of the caliber of Karen Overall - someone certified to work with both meds and behavior.
You might be able to manage this aggression without professional help, but he will not improve. And management means complete isolation from anyone you don't want bitten (including other dogs). Idiopathic means without apparent cause - which means you will not be able to predict when he bites, nor how bad it will be.
With all that said, I really think he sounds like he has a medical problem. What it sounds most like is low thyroid. I had an older dog, my sweet Ben, who was just a teeny bit of a grump around other dogs, at normal times, and a total sweetheart around people. But he suddenly started randomly attacking other dogs, and even snapped at me once. He had a serious medical problem at the same time and the tests we ran for that, at that time, showed the low thyroid. As I said, he was actually "low normal" but thankfully my vet said the symptoms were more important than the test numbers and sure enough, the thyroid replacement got him back to normal in just a few days.
Just last year, a lady brought a Border Collie up to my place for evaluation. She was practically in tears as she described how her dog had started randomly attacking her other dogs, visitors to her farm, and her husband even. I looked at the dog while she was talking and she was overweight, had a really dry faded looking coat, and she was remarkably uninterested in everything around her (my old farm was kind of like Doggy Disneyland with ducks walking around, the ponds, the huge fields, the sheep everywhere, and lots of doggy smells).
She had brought her dog to try her on sheep, and the dog was sort of interested but I could tell she really didn't feel that great. So, when we were saying our good-byes, I told her some things to help manage the situation for a while, said I'd be fine taking her dog after a certain date (I had a dog leaving in a few weeks), and then I told her that the first thing she needed to do was check that thyroid.
She was dubious but bless her heart, she actually did it the very next day. And she wrote me very excitedly to say that her vet during the visit was inclined to diagnose her dog with thyroid even without the test results (that took a few days to get back). Good vet! And she was right - it was the usual "low normal" that BCs have. She wrote a week later to say that her dog was almost back to normal - she was over the moon!
I very much hope that Roscoe has as happy a conclusion to his story.