That's a huge topic to cover here, but I'll try to answer.
The idea is actually rendering components without declaring them inside any module.
Why would we wanna do something like that?
It simple - Modules are much more than just components. Modules have zones, providers, injectors, DI, and much more. Modules for a lot of us represent applications. And sometimes we just want to create a simple component and render it in another component.
What's the problem it will cause?
Modules are the one who sets up zones for us. Zones are the ones who trigger change detection automatically. If we will render a component outside a module, we won't have automatic change detection.
So, with Ivy, we have a few new APIS that can help us:
ɵrenderComponent() - That can render a component without declaring it in a module.
ɵdetectChanges(); - To trigger change detection manually, but, it's just a function from @angular/core and you don't need the DI any more to inject the ChangeDetectorRef
ɵmarkDirty() - To mark the component to check in the next change detection cycle.
ɵɵdirectiveInject() - Inject an InjectionToken in a matter of function, without using the constructor.
If you asking what is this ɵ sign that prefix all those new APIs, that means those functions are still experimental and you shouldn't use them for production yet. And that's also why they are not documented.
For your question - if you want to minimize the use of CD in your components, just render them with renderComponent function, and handle CD by yourself.
If you want to read more, I wrote a complete blog post exactly about this topic, including a lot of code examples. You can find it here - "The future of standalone components in post Ivy release days"
I also gave a talk about it in NG-DE 2019 - "Bye Bye NgModules"
