While the largest fleet I have observed under a single hub managed was fewer than 500 clusters, reviewing those performance results provided good insights.
Thought of sharing here as well since on Linkedin a post by the author on those scale tests went viral (almost 700 likes and 50 plus comments)