Are you required to use Google Earth? I use Google Maps to render a couple thousand points over USA and we use a feature that, depending on the zoom level, it groups those points into markers, and as you zoom in, it will unpack those markers into individual points again.

I am not familiar with Google Earth API so I am not sure if the same is available, but usually when working with that kinda of representation you might want to create a logic or use a feature that will remove the unique points and turn them into a marker, or shape as you mention.

You also have huge benefits for only displaying points that are in an area visible for on the screen. That is also something we do on my implementation of Google Maps. My points are over USA, so if I zoom in Texas, I remove from the GMaps points array any location outside of that zoom area.

For that you will use Gmaps getbounds to be able to get the current visible area and then calculate if your points lat/long are within that area. If you google or look up here in Stackoverflow you will see a couple different ways to implement that.

Also for visualization, you might check other solutions, like Mapbox and OpenStreetMaps. But again, they are nothing like Google Earth.