They added it on Tuesday and removed it today. Fortunately I made a screenshot (from the german version of the page). The support told me exact the same things that has been on the page and the our servers are throttled since >30 days. So it's now hidden but still a fact.
Have you been using 20% sustained for those 30 days?
The surprising thing to me was that you seemed to be saying there was a hard cap of 20% you couldn't ever exceed. I'm not surprised you aren't allowed to use all the shared resources all the time.
A kubernetes master for 9 Nodes on CX21 (old tariff) is using ~25% of memory. So that one was throttled.
Also some of the workers uses >20% on average. Makes no sense to me to book a vServer and not to use it's resources.
If Google translate is correct your link says CPU, not memory. I would be surprised by a memory trottle because I can't think of a sensible way to implement it and for example AWS limits CPU but not memory.
Edit: I uploaded my screenshot: https://i.postimg.cc/dQrwPWdq/signal-2024-09-11-155709-002.p...