There's no vacuum, there's just expanding and contracting streamtubes. At some point you have probably read (or worse: been told) something about wingtip vortices and lift which was at best misleading, at worst false, but has established itself as a great way to sound profound. Essentially the continuation of the unholy "the suction side is longer therefore the air must travel faster to get to its appointment" debacle, which still exists in school textbooks, to this day.
The issue with the video you've posted is that the test section is fairly narrow, which means that the influence of the walls is very strong, and the downwash from the profile is straightened out very quickly. You can of course see that directly on a profile at incidence, the flow is not parallel to the wall, so why would it always have to be strictly parallel behind it?
Look at the video at 8:14. See the downwash?
Here's a few better pictures, probably taken in a somewhat larger tunnel:
You can nicely observe several things:
- The stagnation streamline is diverted upwards, towards the profile
- On the profile, the streamlines follow the profile shape, except in the third picture, which is where it separates
- Behind the profile, due to preservation of momentum, the streamlines cannot just turn a sharp corner and become parallel to the walls -- they flow off roughly parallel to the profile trailing edge.
- The further above or below the profile you get, the more the streamline shapes become smoother and straighter. Once you reach the walls (which are not in the picture), they will be completely straight.
Let's focus on the midle picture because that's most representative of a wing profile in "normal" operation. On top of the wing, close to the tip, where the profile diverts air upwards, the stramlines are "squeezed" together. That's because they're accelerating over the profile, thus narrowing. This means the same mass goes through a smaller cross-section. The further up you go, the closer the spacing gets to "normal", which is because the influence of the profile is diminishing. Below the profile, close to the stagnation point, you can see the inverse: Streamtraces are spaced further apart. That's because they're slowing down. Again, moving from the profile to the wall, you can see the effect fading. Same thing at the trailing edge: On the top side, the flow is now slowing down, and you can see the spacing increasing, while it's acelerating at the bottom. That's partially because of the profile shape, but it's a little exagerated in a wind tunnel because the channel between the top side and the top wall is widening. If the top wall was not there, the streamlines would follow the profile a little closer, but the difference is not huge.
Now, behind the profile, you can clearly see the flow still having a downward component, and the stramlines below the profile still being squeezed a little further. This is because yes of course the profile has been turning the flow, and it can't just change direction like that, even though the wind tunnel walls of course do turn it back to parallel. But there's no need for vacuum to explain that, just streamtubes expanding and contracting, with matching changes to flow velocity and pressure.
I'm not happy with how few videos and pictures of more modern tests are freely available.
See here for a picture of a more recent set-up, and notice how much larger the test section is than the wing profile. This is already much better, but many facilities actually use slotted walls, where some flow can exit and enter the test section, so the streamlines at the walls don't even have to be completely straight.
The maths for anyone who heard about potential flow, with words
This whole thing can also be mathematically modelled, of course. I'm not going to explain the maths behind potential flow models here, although I suspect that the idea of downwash being impossible may have arisen from some attempt at using potential flow to explain lift. For that reason, and in case you have some idea about potential flows: If (in 2D) you overlay a dipole with a constant flow, you get a cilynder, and if you add a vortex at the center, it rotates, creating lift, including upwash in front and downwash behind itself. With enough dipoles and vortices, you can create any shape of airfoil, in case a cylinder doesn't look realistic enough.
Now, if overlay any flow field with its mirror image on any particular plane, there is no flow across that plane, making it a wall. If you generate an infinite mirror cabinet, you can make two parallel wind tunnel walls parallel to the main flow, and your potential flow around the profile in the middle still has downwash -- it just decays towards the walls. No magic needed, no vacuum either.