I know everyone thinks differently and finds different things helpful, but I have really hard time understanding how graph view can be used and what it brings to the table, so please help me understand 
Is graph view part of you workflow?
If it is, please explain how you use it below.
This is a silly use but it’s kind of practical at the same time. Outside of using it in fictional writing, I will watch a lot of movies or shows with friends and I track what movies I’ve watched and more specifically who I have watched them with. Anything to help visualize those connections and be able to move through them.
Graph view as it is right now is a stub feature. In order for it be useful it needs a lot of bread and butter stuff added to it. Filtering by types, filtering by relations, styling according to various criteria, options for controlling layout, pattern matching, named view saving and restoring, and on and on and on.
When you add those things, graph view gets a ton of possible use cases. It can help you spot inconsistencies in your structure, it can help you visualize how closely connected things are, it can help you generate ideas, it can help you create diagrams.
There are more sophisticated features that would increase the value further. With a log of changes, Anytype could show your graph evolve over time, giving you insights into your process. Plus real graph analysis tools could make Gephi unneeded in many instances, and open the doorway to non-academics and non-professionals.
But right now, the graph view is almost completely useless. Its main purposes are to look pretty so our imaginations are fired up, and (educated guess) for the development team to get the force graph library up and running nice and stable within the application.
This subject comes up frequently in the Community, and you can find similar discussions scattered throughout. Here’s one for example: Some thoughts on the new design of the Graph View
Since there are a few bugs with global search e.g. bug with task in search. Graph view is what you can rely on to find an objects ← This is the largest part of my current workflow.
Other than that, I am still eagerly waiting for local graph view for a smooth workflow to trace nearby objects or clutter of objects.
Generally speaking, graph view is good for observation of patterns, and incoming and outgoing behaviours. Distanced individual nodes help linking back to the core knowledge.
The following discussion also has rich contents about graph view.
In Anytype, the graph is created as a core element (user element at the bottom centre, start element) and I think it should be easy to use. I also realise that everyone has different expectations and purposes for using Anytype. Therefore, this is just a suggestion for a possibility.
For a clear arrangement of objects, you can keep your e.g. four burners or most important areas as favourites in the sidebar and not link them to each other. The areas should also be linked in levels that are not too deep. In the end, there are sets or collections.
The advantage is an appealing view with clear groupings of a slowly growing sky with formations of zodiac signs, which in this form also provides new insights. It is then not a ball of wool or a cloud with overlapping links.
From my experience with Obsidian (incl. Dataview) and the Dailies First method, the existing graph is an unusable chaos. Certainly, tendencies can be recognised through condensations.