I’m still investigating.
I did a test on my Android device (Android 11). it also crashes there and kills my browser.
On Edge on my desktop PC. It works but it allocates up to 4Gb of ram.
I wasn’t expecting all of the files to be loaded to the scene or kept in memory using the .addAllToScene and .removeallFromScene I assumed only one would be in memory. Because the draco encoded glb files explode in size when they are decompressed. I uploaded a 4Mb draco file to the sandbox and exported it as glb it is 52MB
Looking at your scene, there are meshes with a huge number of faces. I think there are areas of improvement on babylon side but your meshes are too big to fit in a browser memory.
We will do some test to improve caching and memory, this will take take. But it won’t be enough to properly handle your scene.
Thanks for your help on identifying the major issues. Would you take a look at this PG? It is working locally on my phone, it would just be nice to have confirmation that it works for others and that the memory usage stays low (chrome says it increases during each load even though its from the cache, but task manager shows the memory usage is stable).
I used an xhr to grab all the files first and then only open them one at a time, rather than having all the files decompressed in ram.
I am running into an issue given the async nature of the load, if you scroll too quickly on the slider things fall apart (trying to sort that out)