Questions regarding Fleet and design decisions

I’ve informed myself about Fleet and the “Cluster on every toaster” philiosophy.
However, I am still not entirely sure, why exactly it is better to have lots of clusters under the supervision Fleet provides instead of a huge cluster with a lot of nodes.
I know that it relates to connectivity issues and that lack of network connectivity is for example supposed to be easier to handle on a cluster level rather than on a worker node level. Though, is this still not just moving the goal post? What if Fleet fails?

In what way is a cluster level management precisely and indisputably better than a worker node level management design philosophy?

Thank you for any insights!