I am working with a regular mesh of points in an app developed with Web AppBuilder. This dataset has tens of millions of points and the geographical scope of work is very extensive so I have to see the data set at E 1: 1,000,000 for a complete overview.
Attached image shows a sample of the data at a scale of 1: 5,000 (red box), and 1: 1,000,000 (blue box).
I have been working with this  ESRI project sample to vary point sizes by scale, but my dataset is too big and refreshing screen is very slow.
So, I would like to reduce the quantity of points (generalization ) from E1,000,000 to E1: 10,000 because they were used only for visualization.
A simple technique of generalizing points in regular mesh is to construct a mesh subset of a mesh, and so on for the different scales.
Here I have two doubts,
1. For a given map scale, which mesh pitch is appropriate? Is there any theory about it?
2. How can you generate a mesh from a subset of data taking into account the values of the points of the original mesh? I have to preserve the symbology of the points on all scales.
Any suggestion or examples will be greatly appreciated.
Thank you in advance!