I've not been inside my pump but I'm not afraid to do it if I have to. You should be posting photos of what you are doing, it helps others even of there's no obvious conclusion.
What I do know is there are 3(?) electric solenoid valves controlled by the roof controller which steer the hydraulic fluid according to what is required in the roof cycle. I don't know where they are, but if the roof pump itself has more than 4 wires (power+thermal sensor) then that is where they will be. If vagcom shows all the sensors operating ok and there are no bent or distorted parts, I think roof faults will boil down to 2 parts - The roof controller (electronic control) or the pump, hydraulics and valves.
One thing I'm curious about is there is a fluid reservoir, but once you start disconnecting lines there is a likely hood of air getting into them. If this gets into the rams they will not move with the same force. My curiosity is therefore how do you bleed the system if a ram has an air lock and faults out the roof to stop? I will have a look through my resources but I'm sure I haven't seen anything yet. I have some ideas that you could push fluid through a pair with a hypodermic syringe and quickly connect the pair to the pump. You could probably momentarily run the pump by connecting power to it with the return fluid line open. But that is highly risky because you would need to know the positions of the valves and which line will be the flow and return. These pairs of lines carry fluid in both directions to either push or pull the rams. O.K I have the answer: The hydraulic system is a self-bleeding system. Opening and closing the roof twice bleeds the system. But check the reservoir level.
OK I see a catch 22. Once you start opening lines, if you can't open and close the roof you can't bleed the system. You can expect the roof operation to be 'jerky' if there is trapped air, so you repeat the opening and closing until the hydraulics operate smoothly.
This is the recommended hydraulic oil:
If I was experimenting myself I would test the theory on the trunk lid because once you remove the trim covers you can more easily see what the ram is doing.
The real weakness in all of this is vagcom does no more than give feedback on sensors which limits it as a fault finding tool to do things like start/stop the pump, control valves and override sensors. All these things are really risky with potential to crunch a roof if you don't know what you are doing.
I am working on an idea that since nobody has a software diagnostics tool to take over fine control using CAN controls (and I am not a software writer!) All these things could be done from wiring to the roof controller where you could play your own tunes, independently of the control program. At least you get access to the real world roof events and not a diagnostics error message that stops you going further. At the moment I'm looking into wire taps that might be used that are not too bulky to produce my own 'mimic' panel. Looks like about 20 taps might be needed which I'm not over the moon about doing.
I forgot to add that just for fun I removed the trim covers last week which cover the trunk hinge. Removing the covers without breaking them was itself a challenge! Inside on the far left (and right) is the hydraulic ram responsible for opening and closing the locks, but I think the front top roof member lock is operated by a Bowden cable as the hinge operates. I think the lines were coded 51 and 52 (?) for the left side. This is the ram piston which has to be moved manually (with a plastic taper wedge) to unlock and then manually move roof parts. It has to be the first and last operation of opening and closing the roof. The locking system has status sensors feeding into the roof controller. I don't know why VW use this general purpose body tool apart from avoiding damage to the polished piston ram. It just seems the wrong shape. In your case you could try opening the manual valve on the pump and moving these two ram pistons forwards to the unlock position. Then with 2 people one either side, you can check each of the roof parts is able to move freely. If that is the case then I think that leaves you with a control electronics fault or pump hydraulics.
What I said about making my own real world mimic fault finding panel is true here. The roof controller has various 'output' functions usually done with a strong mosfet inside the controller box. Software can tell you that a particular output function is 'active' but in the real world it won't tell you if the output mosfet has died or a valve is faulty. Using a connector wiring diagram I know enough to tap the relevant wires and test. This won't be for everybody but the example illustrates how you might identify whether the controller or pump or both are faulty. The dealer approach would probably be to replace both and charge a lot of money, because they don't know how to test each component.