It depends. Three adjacent views on the same scene, just projected at different angles and translations, can be rendered as a single pass on the GPU. This is much faster than treating them as individual views and rendering them as separate scenes. That’s what P3D does and the performance impact of multiple views is mitigated substantially. Also, DX12 improved that situation quite a lot over DX11.
Ultimately, though, the FSX / P3D view engine is based on the ‘multiple window’ system that was in Flight Simulator back before Microsoft bought it. The Atari ST version in the early 90s could have multiple windows, each with a different view, displayed at the same time. Yes, with performance loss per-window, of course, and only on a single screen since you couldn’t do multiple monitors. When it got ported to DirectX and the OS permitted multiple displays, the ability to undock a window to another monitor was added. The key thing is that the multi-window capability was complete; you could always interact with controls in any window, for example. It could be used to spread 2D panels around your monitors, while still allowing you to interact with them. It could be used to chop up the outside view into parts and push them across an arrangement of monitors, or, with other software, to create a wide-FOV view on multiple projectors. That system was the key. MSFS doesn’t have that, it has non-functional pop-ups which are just copies of the texture being rendered in the VC, which you can’t interact with. That’s what’s missing, as a minimum. P3D also added direct control of the view frustums for each view, and that would be important to have too for advanced display configurations.
Then we need full camera control via SimConnect so add-ons can control the camera. That would be true parity with FSX / P3D and that IMHO is the minimum we should be expecting and asking for. None of it is unachievable.
Once you have all that, a fly-by view is child’s play to implement.