It seems that it is a hypothetical bystander that must fear for their safety, as opposed to actually someone being there so I guess the evidence of the phone videwitness is important.
Yep. Its a public order offence, so its concerned only with the actions in their relation to the general public. The protagonists are committing a crime against the public good, not to each other. If these actions happen in a situation where, hypothetically, a random person could stumbled across it, then you have the possibility of an offence. In the most simple terms, if it happens on the street like this did, all the available evidence will be related to how that invisible person would react. Not how people present actually did react. If a group of I dont know, ex-Afghan vets who were 6ft 5inch stood watching laughing at them because they have, to some extent, been desensitized to violence, then it wouldnt matter. The possibility is, at any moment another random person could turn up, so why does it matter who was there. There is one other person at the party all the time, and it will be average-joe, invisible and not existing, but hes put there standing next to it at the front row..... its him who provides the test.
However surely the defence will try to argue that the violence was solely between the three up on the charge of affray, and they weren't looking to bring any innocent bystander into the situation.
A natural conclusion, but the problem is if its in the public view, and he can hypothetically be there, he will be. And he will see everything that can be proven to take place, and will react to that in the good old average way someone will do.
The really struggle in the trial will be the prosecution hammering home as many examples that can of actual people, because it is far easier to believe the average person would be scared, if you can demonstrate the average person was scared. The defence really have to find a way to separate being scared or shocked, and that turning into a fear of violence towards that person (which I accept is very close to what you were implying).
That can take many forms, and here lies the problem with predicting a case like this; while its all objective and hypothetical, the jury in the end will probably make a subjective judgement as if they were scared, if they would fear violence. And its far easier to envision something being relatively safe when your 12 months after the event, in a courtroom. The lady in her house was in her house, was she in any real danger. So the human element of this plays out, the lawyers and the judges can guide and tell the jury their opinion as to the law, how they should decide, what they should consider; but its impossible not for these people not to decide on their own feelings, guided a lot by the experiences of real people. Its hard to take subjectivity out of it and place conflicting emotions onto a hypothetical person.