For me, we are concerned with designing our plants to ensure humans can interact efficiently & safely with the systems (power plants) we're designing. Here are some benefits for me of VR over 2D viewing of 3D models:
1. Visibility assessments. In VR you can get a true sense of Field of View (FOV). For example, operators want maximum visibility through windows in control stations (or in a flight control tower, or sitting in the driver seat of a car). In 2D viewing, you can adjust the perspective settings to be anything and you simply can't match a humans natural FOV, which in addition to matching true perspective, you also capture the real time dynamics of how visibility changes as a user's body/head move. I've seen control stations completely redesigned after being built because the customer didn't like the experience (a risk largely mitigated with VR). You can't put 'control stations' in glass bubbles. It just isn't feasible. You need reasonably-sized windows and mullions, and there are many obstructions to contend with. 2D visualizations of 3D walk-through models simply don't accurately capture what's a VR experience can. We've actually found that, concerns about visibility identified from 2D visualizations are often discounted when we check it out in VR, just because the way and how quickly people move around naturally and subconsciously improves visibility.
2.) Ergonomics (Human-Machine Interface). In VR, you can literally reach out with your physical hands and determine if a valve, or a button is safely reachable by an operator. In initial design, you rely on standards/rules of thumb for placement of these items. However, as the design matures, you inevitably end up with pipes/wires/beams that complicate matters - it's impossible to right rules for every possible scenario. Historically, the only way to be sure.
3. Efficiency. I've found that user's can intuitively move around in a VR model much faster than they can with a screen/mouse. Especially, when you have a complex nest of pipes you're trying to evaluate. Using mouse/joystick/keyboard clicks and rotating takes people much longer than simply bending down and turning their head to see an object hiding behind a pipe or other object.
4. Remote Inspections. Once the plant (or whatever) is built, it's typically not near the office of the engineering firm that designed it or customer's headquarters that paid for it. Historically, when you get reports of people complaining about a design, you go into a 2D model first to check it out. If you can't get a good sense of the issue (typically because you can't get a true feel for the ergonomics/visibility aspects of the issue), you buy a plane ticket, rental car, hotel, and go an inspect in the plant in person. However, VR is good as assessing ergonomics and visibility issues (as described above), so you can use VR to avoid the need for travel. This has occurred already on a few occasions.
We currently only use static models, but smarter people can begin encoding dynamic features (e.g. opening doors, removing panels, etc.). I'm sure other people are doing this already. In conjunction with the ergonomics/visibility aspects above, I expect this will greatly enhance VR benefits.