That's the problem. I think if you define "educational" as the curriculum taught in school, this is not possible. There's no way around working a ton of math problems to get better at math. You maybe can make it a bit more engaging if you pack it in a video game, but it will never be fun for most people.
Therefore, I think, the educational role of video games must be to spark interest. For example, KSP sparks interest in aerospace engineering, and I bet there's a significant number of people who went on to study rocketry because they played KSP.
Sparking interest is an important function. In fact, I think it becomes increasingly important as a lot of material and courses are now available for little money ubiquitously. Great education starts with a spark of interest.
Let me explain in more detail:
Back in the days, an important part of going to school was having access to knowledge. Now, thanks to the internet and millions of amazing humans, knowledge is accessible all over the place, virtually for no money. Today, the motivation to learn something is therefore more important. To get motivated, first you need to know that something exists, and second you need to know if you like it or not. Before motivation comes sparking interest. And here, video games have an extremely important role to play - and unlike "educational" video games that are math exercises disguised as video games, video games that can try to achieve the spark can be genuinely fun to play.
About 'German Grammar': check out Heaven's Vault https://en.wikipedia.org/wiki/Heaven%27s_Vault
But honestly, humans don't learn languages by understand their grammars with the logic part of their brain. Games would be ideal to produce comprehensible input, and test players on their comprehension via actions, instead of making them reply with words.
There's some evidence that trying to produce language to early in your learning just ingrains bad habits. So instead you can just follow increasingly complex instructions to show that you understand. See https://en.wikipedia.org/wiki/Input_hypothesis
As a native German speaker, I disagree. For me, the German grammar (in particular the declination system) is what I would consider the type system of the German language. So yes, I do think a lot about German sentences in terms of types/grammar that are/is involved.
EDIT: When I was a pupil at school, the only thing that I loved about the German classes (for native speakers) were the grammatical concepts. At that time, I really couldn't understand why these weren't taught in math classes ...
But you first learn a language intuitively at home, by using it a lot and developing intuition for what sounds right and what doesn't.
I'm a native Polish speaker (which has a "typesystem" much more complicated than the German one nevermind the English one). I've learnt German for 4 years in secondary school. I remember the system to be pretty intuitive and simple (mostly you change the articles not the words, there's just 4 cases, no verb aspects, just 1 kind of plural, just 3 genders, etc.).
So it was definitely easy to remember the rules and I did at one point (not anymore ;) ). But still to speak German you can't pause every 3 seconds to do a table lookup for the correct conjugation and rules, that would be far too slow. You have to use the language and have that "ouch" feeling when you use the wrong case so eventually you know intuitively which word to use.
A game to teach a language should just be a game using that language and giving immediate feedback.
Learning the first language as a baby takes many years - and even after these years, you are still only on a "baby talk level". That is why I don't consider these "natural" approaches for learning a language to be a good idea - they are far too slow to be economical.
WWII is a pretty common setting for games. Making the German protagonists communicating only in German would be a first step to incentive learning the language.
Then mix and match for different languages.
Of course the things they say should be more detailed and diverse than just some standard sound clips.
And that's the problem. Because they are educational. And because they are both educational and fun, people are much more engaged to learn from them than from "real" educational games.
People who play Europe Universalis learn way more about early modern history than they ever learned at school. As Randal Munroe pointed out, you learn way more about orbital mechanics from KSP.
If you want to make truly educational games, you shouldn't focus on the educational part, but take the educational part and wrap it in tons of fun.
Of course there are topics where this is going to be hard. I have absolutely no idea how anyone could make a fun game that revolves around German grammar. (Or do I? The best way to learn a language always seems to be to actually use it with a native speaker. Having a friend who speaks the language you're trying to learn would be a great way to do that. There might be something here.)
But something like geology could be part of a simulation where you need to find certain resources, and that's easier once you understand how those resources are formed. And then there needs to be something fun to do with those resources, of course.