These are opposed. A more decentralized staking set is one that can run on a cell phone (which PBS achieves, and Ethereum may).
I’m concerned that the centralizing effects of such high operating costs (capital/electricity/bandwidth) are being underestimated. I do not expect stakeholders to want to run their own hardware, and to readily stake via third parties instead.
This creates a dependency on a trusted third party (relays). PBS addresses this issue.
You’re confusing proof-time with block-time. Both Fernet and PBS can produce blocks at ~L1 speeds.
This also implies that multiple concurrent committees must be electable (requiring ~10x nodes given 12s blocks and 10min proof time).
Provers may be cheaply bribed not to prove in order to reorg the chain. Because there is no data-availability guarantee, we cannot slash provers who don’t participate. This may cost less than prover-block-reward to do, and will also cause the sequencer to get slashed.
Provers may collude against sequencers, because the sequencers stake is at risk if the provers refuse to sign.