graphics card - Nvidia Optimus not connected to any display?

While trying to spec out a Lenovo T440p ThinkPad with Nvidia graphics and a docking station allowing for dual monitors, I found a forum post implying that the Nvidia Optimus could not be used with docking station outputs. This claim was made from display topology information found in a video driver's README file:

Since Optimus Technology has the ability to use two display adapters, some display monitors (or displays) are assigned to specific adapter. Below describes which display adapter controls the display connector for a certain display monitor.

Thinkpad T440, T440s, T440p, T540p

 Intel HD Graphics
     - (Computer's LCD)
     - Computer's analog VGA connector
     - Computer's DisplayPort connector
     - Docking Station's analog VGA connector
     - Docking Station's DVI connector(s) 
     - Docking Station's DisplayPort connector(s)
     - Docking Station's HDMI connector
 NVIDIA GeForce GT 730M, GeForce GT 720M
     - No display is connected to this display adapter.

What good is the Nvidia adapter if it isn't connected to any display? I assume there is more to the story here, but I am not familiar with how these two graphics adapters interact with each other. Is there any way to use the Nvidia adapter with docking station ports? And if not, how does the Nvidia adapter work at all?

The text file goes on to talk about "standard mode" and "advanced mode". The "advanced mode" seems to allow more display pin configurations, but isn't available on the T440p.

There is also a note about cross-adapter display cloning, but I do not believe this applies to my situation.

Answer

Imagine that you have your screens physically connected to the dedicated/discrete NVIDIA graphics card. When switching to integrated Intel GPU, the display would have to be literally disconnected from dedicated GPU and connected to integrated one. That would lead to a number of undesired effects (for example screen would be black for a second or two before second GPU boots up and recognizes it) and would be hard to implement reliably in hardware.

Instead, all displays are connected to Intel integrated GPU that is always powered and running. When more powerful GPU is needed NVIDIA kicks in, but integrated graphics are still on. Dedicated GPU only renders frames and sends them to the integrated one, which only displays them on the screen without any processing. One drawback of this approach is that framerate is limited by integrated GPU, but it's not a real problem because it's only displaying still images sent by discrete unit. Intel's GPU can easily reach few hundreds of 1080p images per second, so it doesn't really affect performance in any way.