I don't understand what is wrong.

I listen to a track from Michel Portal, Mutinerie, from Dockings.
This is left/right signal as usual. Screenshot at 01:00.

Then i generate a file, same take, with middle/side signal.

Top, grey: left/right signal
Bottom, blue: middle/side signal
Switch the Vector Audio Scope [Decode M/S] to ON.
Listen to the modified take.

It's not easy to hit the same time exactly.
But the orientations of the Vector Audio Scope are the same.
Music mostly in the center is vertikal.

To keep it as simple as possible, assuming no correction for different microphones and amplifiers, the following applies
L = M/2 + S/2
R = M/2 - S/2
or, the other way around
M = L/2 + R/2
S = L/2 - R/2 
Mathematically, this is a rotation of 45°.

You also switch on X/Y mode. This is a gimmick that people use to create digital audio files that are not for listening, but for drawing nice graphics in a Vector Audio Scope.  To ensure that the coordinates +1/+1 are at the top right and -1/-1 at the bottom left, a 45% rotation is required, because normally in the Vector Audio Scope, correlation = +1 (+1/+1 to -1/-1, S=0) is a vertical line. Both switches On at the same time means 2 x 45°, that's 90°.

M/S and L/R are both 2 channels, stereo. For L/R stereo, channel 1 is usually L and channel 2 is R, see Input Selection.  Except if you have a microphone setup using the M/S technique, then channel 1 is M and channel 2 is S. However, there is no way for software to recognize which microphone setup the sound engineer uses. The sound engineer must tell the software this itself; this is what the Decode M/S switch is for. It is not an Encode M/S switch.

In my humble opinion, there are two possibilities.
On the one hand: Your source software calculates the M/S signal on the fly and delivers 4 channels.
Input first Vector Audio Scope
L = 1
R = 2
Setting Decode M/S = Off
Input second Vector Audio Scope
L = 3
R = 4
Setting Decode M/S = On
On the other hand:
RME adds a new function: Encode M/S. Then both Vector Audio Scopes would display the same thing. For example, correlation = +1 (L=R and S=0 ) as a vertical line.

In both cases: what's the point of having the same graph twice, just calculated differently? In your setup both Vector Audio Scopes show exactly the same figure, only rotated by 90°.

And why do you switch on X/Y mode when viewing real audio recordings at all?

I still don't understand the request.

I do not understand, what probem do you have.

If Mid is stronger than Side then it is correct that the Scope pattern is horizontal. What else would you expect? In case Mid would be weaker then Side, the Scope pattern would be vertical.

I haven't forgotten that. I played back an existing recording because I don't have a bidirectional microphone. Generating the file was my "post" production.

The [Decode M/S] option is there to ensure that, when M is on channel 1 and S is on channel 2, the center is vertical as usual. Only then does it make sense to enable the option; during recording, post production and listening.

If you send the two signals to a normal oscilloscope, L to the horizontal axis and R to the vertical axis, it shows a diagonal line when L = R, i.e., correlation = 1.

In the case of an M/S arrangement, however, M = vertical axis and S = horizontal axis, a vertical line is created from the center (M changes, S remains 0).

As GolfPutter already wrote, this fits better and is more intuitive. What comes from the center is displayed as coming from above.
That's why, in Audio Vector Scopes, the oscilloscope tube was rotated by 45° (before the digital age), which was easy to do mechanically, but is stupid in a digital device with a pixel screen. It is easier to leave the screen normal and calculate the rotation.

If you switch Digicheck setting [Decode M/S] to On, the graphic rotates 45° back.

If you absolutely want the classic oscilloscope display, for whatever reason, you can switch on [X/Y-Mode]. DigiCheck then sends L to the horizontal axis and R to the vertical axis. If you enable [X/Y-Mode] too, it rotates another 45° more.

If RME would fulfill the above request to turn this up, the result would be:

Absolutely identical to L/R Audio Vector Scope (see above). In my opinion, makes not really sense. If 2 Vector Audio Scopes do display different shapes from the same input, it's because other options in the Audio Vector Scope Settings are different.

In the first screenshot from GolfPutter the shapes are equal:

Only the calculation of the correlation -0.05 goes wrong, because Digicheck uses the formula for M/S input, but the input is L/R.

As an aside: Vinyl stereo records are actually also M/S arrangements, only rotated: M is horizontal, S is vertical. This is because mono came first and was recorded horizontally. But in the turntable's stereo pickup the coils are already arranged at 45° angles, so that it directly outputs L and an R signals. This also applies to many radio signals, so that old or simple mono receivers continue to work.