I’ve been doing some research for my own stuff.
And while you can build a camera rig using BP, and have two cameras with user definable Eye Separation and Convergence, I have a working prototype that works great except for anything like particles, or any animated displacements like simple grass wind etc where it is always simulating in the BG, and therefor rendering each camera will yield different results.
Now you can render these camera views to a render target and film that render target with an Ortho-camera as some others have done in the past to get a synced side by side render. However for me this is an unacceptable method, not just do to performance reasons (expect UE4 to come to a crawl at any usable resolutions) However you do not get the benefit from things like motion blur, and several other post process rendering features. I also found that Anti-Aliasing also suffered a great deal so I would have to at the very least render at a much higher rez and down sample in post. The biggest issue however is forward looking. By forcing me to have a side by side render. I can’t do things like integration of UE4s renderer with other renderers in other apps. This may be a small use case in the grand scheme of things, however for me, as an offline VFX artist, multi-pass rendering is actually quite a big deal. Not to mention the ability to A\B my cameras to see how the eye separation effects the distance between the cameras in a very visual way.
So at this point the solution has to be a C++ solution, since you will always have to capture each eye at each frame in the simulation.