I think Ethereum 2.0 is currently demonstrating at scale that proof of stake works fine and can be secure against hordes of very incentivized hackers trying to cheat. That should hopefully bury the whole proof of stake is centralized argument.
The notion of centralization has its roots in a different concept: permissioned vs. permissionless blockchains. Ethereum (1 and 2) and Bitcoin are both permissionless in the sense that anyone can become a node in the network and start earning gas. Ethereum 2.0 does not change that. You can still join and earn gas on it.
Permissioned blockchains such as Stellar, Ripple and a few other networks are more closed. You can join the network but you don't have the same status as the core nodes in the network that are explicitly configured to trust each other. This makes e.g. 50% attacks from the outside impossible. E.g. Stellar works this way. You can become a stellar validator but nodes have to specify which other nodes they trust and that gives the first nodes a lot of power.
In the case of Stellar there are only a few dozen validators run by organizations working with the Stellar foundation. Stellar has indicated that they want to gradually open up and eventually become permissionless. But as of yet they are too small to make that work and are relying a lot on a small set of validators run by just a handful of companies. There's a list of active nodes here: https://stellarbeat.io/nodes
Such permissioned networks are great for e.g. banks that need to make sure they aren't doing transactions via some dodgy nodes in North Korea, Iran, etc. Technically, blockchains like this are not that different from public blockchains. There's nothing stopping these from opening up except that they choose not to. For a lot of businesses, that's a feature not a bug.