In my experience, these "interviews" are not like formal job interviews. I do not think, in physics/astronomy, the interview will consist of questions to test your physics knowledge or background.
Instead, they are there for you and the interviewer to get more information about both parties. My "interview" began with the interviewers telling me about their program -- I didn't even speak for the first 10 minutes. They told me about how long the program is, what exams I would have to take if I'm admitted, what the department's focus is, how many students they have etc. all that. They even talked a little about their own research even though I didn't mention them at all in my application!
Then, the conversation turned towards me and they asked me about things you would normally put in a SOP. I talked about the classes I took, what research I've done etc. The interview is not just a repetition of the SOP though, since you are able to speak a lot more than you can write and there is real-time interaction to clarify details. I suppose this is also a "test" to make sure you truly understood your research project. However, I felt it was very "low pressure", since they were not grilling me like my thesis defenses. Instead, it was more like if you were talking to a colleague or someone you just met at a conference and they ask "so what do you study?". By this I mean that they didn't ask you things they already knew the answer to (as if to check my knowledge). They only asked questions that they wanted to know the answer to, like what the main result of my project was and why people care.
The interview procedure varies from school to school. For the interview I had, they said that they were interviewing everyone that they had short-listed, not just like "borderline" people. It might be the case that the type of interview you get depends on where you rank in their shortlist, but the overall impression I get from talking to people is that "interviews" in physics/astronomy are not like oral exams.
I would prepare for it the same way you would prepare yourself to talk about your project at a conference -- except the topic is your entire academic/research history instead of just one project you're presenting. In my opinion, it is VERY useful to practice a short description of each project you have done. Identify the key research question and your main result. Also identify what parts are important to share and which details can be left out. For example, you might want to say that you use Method X to calculate result Y but maybe it's too much in the first pass to explain every detail of Method X. Or, if the details are important (maybe you invented/used a new thing) then decide (and practice) ahead of time. For me, when I try to explain science "on the spot", the narrative can get convoluted and I might get trapped in the details or leave out something important. Maybe even practice with a classmate and see if they can understand a 2 minute "pitch" of your project(s)!