Bosch VE fuel flow path.
The fuel pump inside the IP is a rotary vane pump on the input shaft. The vanes don't have springs, and are held out in contact with the eccentric ring by centrifugal force. That means it won't pump unless it's spinning rather fast. It has a pressure regulator that recycles excess fuel back to the inlet.
The pumped fuel simply dumps into the IP interior (the technical term for this large volume is an "Ullage" volume). The pressure regulator works to maintain a nearly constant interior pressure.
At the highest point on the IP is the overflow restriction. Any air bubbles eventually find their way out of it. It's more than a bolt with a precise hole drilled in it - it also has a perforated steel tube that also affects pressure at different flow rates, so the whole thing is calibrated.
The high pressure plunger that actually generates the pulses of fuel for the injectors gets it's fuel from a port about half-way above the bottom of the IP.
The spinning parts inside the IP are like a blender, mixing freshly arriving fuel with fuel that arrived some time before, while simultaneously bleeding out some of the blended fuel(s), while also pumping some of it to the injectors. This is important for SVO operation because the time needed to "purge" the system is longer than would first appear. The percentage of SVO remaining inside the IP (and therefore being fed to the injectors) reduces slowly, even though lots of fuel is flowing through. Because of the violent mixing that takes place, the fuel inside the IP is homogenized, so every time one volume of fuel enters, and one volume exits, only half of the departing fuel will be from fuel that was already there. The other half departing is the new fuel.
Here's an example: think of a 2-quart blender filled with, um, milk (to simulate SVO). Turn it on low, simulating the spinning parts inside the VW IP. Now slowly add water (simulating diesel), letting the displaced milk run down the outside. How much water must be added before all of the milk is gone from the container? If you say "2-quarts", you're wrong. When you've added 2 quarts, half the milk will still be in the blender. Adding another 2 quarts of water displaces half of that (25%). Another 2 quarts makes it half of that (12%). You can see the problem - there will always be some % milk left, but it gets to be a very small %. If this were SVO and diesel in an IP, how much residual SVO is acceptable? That's a very good question, but I haven't got a scientific answer.