The notion that “it is practically manageable” is not really correct and it just gives a false sense of security. Note that Hedera requires 2/3rds of total money to agree, which is not the same as 2/3rds of nodes which is the byzantine agreement requirement. 2/3rds of money can be located in any number of nodes from 1 to 1 million. Similarly 1/3rd of money can be theoretically located in one node or a bunch of people who will act like one node. When they shut down the node, whole of Hedera will stall. This is akin to your apprehension that even 1/3rd of the total money may not be out of hands of some rich folks.

Note that Algorand does a bit better by allowing a small random subset of nodes to be the validators. But that has its own problems.

The root cause of the difficulties with these protocols is to use PoS as a validation mechanism on top of BFT. It confuses two domains of economics and technology together, which results in an unholy mess.

Have a look at my TrustChain protocol, where I try to make it a bit better than Algorand and Hedera. Still a work in progress…