Outputting pixels of "completed work" is generally a low-intensity task of modern GPUs.
You have dedicated hardware that is responsible for video output, pulling data from framebuffers and putting it onto external displays.
In terms of everyday work—not including 3D applications or video editing initially—the main use of VRAM would be the framebuffers, textures for UI elements, and any additional buffers needed for GPU-based operations like window compositing and transparency effects, etc.
A single frame on an Apple 5K display is around 56MB and macOS uses double buffering which makes it 112MB. It's safe to say that the system would require at least a few hundred megabytes beyond that, likely edging closer to 1 GB or more in a typical macOS environment with multiple applications open. And yes, that's just a "what the hell, let's just call it 1GB"-number.
While not a negligible number, it's manageable. On top of that, you have all the computational work that you do on an app-per-app basis.
Still, as you hopefully understand, adding an extra display is not too much to ask of today's hardware.