So, I ended up implementing something like what @Jahnold recommended. (I will post the diagram in the link provided for the idea of fooobar.com/questions/1265735 / ... )
Hannes Dorfmann (the guy who created / managed the famous MVB Mosby library: Github link ) also pointed me in that direction.

Implementation
I have a presenter for the main activity and several fragments that can be used in this activity. Each fragment has its own leader. Then I use the repository (search for the repository template), which basically stores the models and business logic. For my use case, I save this repository as a singleton. The repository provides data in three forms: from an online api, sqlite database, or cache stored in memory (mainly for arraylist elements). I also have some currentitem int indexes and stuff in this repository that are updated depending on the current state.
Therefore, data, state, and business logic are stored in this shared repository. Leading and views are pretty dumb. I donβt have much business logic (application-specific logic) among the presenters. They just have logic related to how data should be displayed (view specific logic) and preprocess them in the logic.
Example
Whenever a fragment and activity should talk to each other (through speakers), when a user presses a button in a child fragment, the fragment requests handleClick from its presenter, the presenters update the current storage data currentItemSelected (or something else) and requests the fragment to run events (for example, onbuttonclick) for an interface listener that implements the action. When an action receives an event, it asks it for its own host to handle it, and in turn, the activity host looks for an update in the repository to get a new currentItemSelected.
Additional information (extended version) :
You can also keep an eye out for clean architecture, which is a kind of more advanced version of the MVP architecture. MVP simply deals with view architecture, where since pure architecture also deals with business logic and data architecture, MVP is just a small part of the clean arch that is used to handle views. Using this, you can break the mega-repo in my case into even more used cases (or interactions) that handle a specific case of using business logic, and the repository simply provides data. Thus, the logical flow is now a view -> presenter -> interactor -> repo and vice versa.