Hi!
I think this scene is still rather a large one when considered for internet delivery. It has ~10M triangles, while we usually suggest to ideally have 2-3M triangles.
If you managed to decrease the number of triangles, the download time of the scene will decrease significantly.
You can find the most complex objects in your scene by analyzing the objects in the Objects tab. The tab gives you information on the number of faces (triangles) in the currently selected object and in the whole scene. The list of objects is sorted starting from the objects with the highest number of faces (at each hierarchy level), so you can identify the objects to replace/optimize by looking at the top of this list.
Then, you can find if there are any less relevant that could be removed without significant visual impact. In particular you can check roughness/metallic/bump textures.