My only real interview (i.e. one that happened before learning about the admission decision) was a Skype interview with U Washington Astro. I don't know if you ever interviewed for research positions (e.g. for undergraduate research positions) in the past, but my experience was like the many other undergrad research interviews.
There were 2 profs on the interview. We spent the first 10-15 minutes just talking about their own research and the U Washington Astro program. They spoke almost exclusively for the first 10-15 minutes, with only me saying things that acknowledged understanding (like "yes" or "cool" etc.) and asking a few small clarification questions.
The second half was more like a regular interview except the questions were not particularly probing. They asked me why I was interested in astronomy, what research topics excited me, and why U Washington specifically. I basically just repeated everything in my SOP and CV, but with more details. Of course, I didn't recite my SOP (this interview took place 12 weeks after the school deadline so I didn't remember it anyways). But my point is that you should not worry about repeating yourself. In fact, it is very unlikely they will even remember the details of your SOP (they might have glanced at it quickly before calling you though). So, when you answer the questions, don't assume they have the SOP well read---answer it as if they are asking you for the first time. (Also, if there are lots of applicants and a large committee, the faculty interviewing you may not all have been assigned your application package in the first stage).
Of course, with that said, you should really be prepared to answer deep questions too. They might be scanning your CV to find topics for questions. So, you should practice 30-second elevator pitches for each of the research project you've done. You might know your own work well enough but I find that interviews are a little scary and having that 30 seconds memorized is a good confidence booster. You should also be able to answer basic questions about your work:
1. What methods did you use? Why did you choose those methods? What other methods are possible and why didn't you choose them?
2. What is the main research question you want to answer? Why is this important?
3. What is the big picture question? Where does your project fit into the field's understanding of this topic?
4. What is your main result? What does it mean? What implications does it have, if any, for the answers to the questions in 2 and 3?
--- If you don't have any results yet, be prepared for questions like "If you did/didn't find X, what does that mean?"
5. What is the next step? What caveats come with your current result and how would you address them?
Overall though, I think the main purpose of the interviews in astro (and maybe also physics, but can't speak for that) is NOT to test/probe your scientific or research ability. Instead, it's a test of your ability to interact as a professional in the field. Yes, it's important that you know the science behind your work (see the sample questions above) but really, I think the most important part is HOW you answer the questions, not what you are saying. They are looking for someone who can think critically and communicate well.
I think that graduate departments aren't necessarily looking for someone to come in with lots of graduate level physics knowledge---that's something they can teach you in grad school. They want students who can show initiative, think for themselves, communicate effectively, take leadership over their projects, etc. That's why the above questions are important: how you answer them can help them determine what kind of role you played in the projects. And these skills are things that physics/astro departments aren't really as good as teaching (after all, they are a research department, not a professional development center), so I think these aspects are more important to grad school than advanced physics knowledge.