Why exactly does immediate-mode GUI rerender the scene and why doesn´t retained-mode?

Question

I guess Casey Muratori formulated it well:

"[...] except to add the fact that if you’re arguing about it, you probably don’t understand it very well."

Maybe not for the arguing part... Well, I´m not trying to argue here, nor do i want to heat up that biiig discussion wether ImGui or rmGui is better. What I seek is some explanation! So far i understood that little problem everybody mentions: ImGuis have to rerender/rebuild (idk how to formulate exactly) their model more often than rmGui? - Questionmark intended because then people argue that it doesn´t and that´s where i get confused.

Could someone explain to me the following:

• Why does the rmGui not need to render the scene every frame? I mean it would have to tell the Computer what to show as well?
• And if rmGui doesn´t need to then why does ImGui need to?

Is it like that rmGui stores what it has to send to the GPU or whatever and by that only needs to change depending on action? But why can´t ImGui do that too like keeping everything the same until some action?

Thanks for reading - I read posts for the past 5 days and I still don´t really get it all. Reference would be realllyyy nice then I could dig a bit more into the topic.