I keep hearing that, and yet it doesn't happen. Did anyone ever actually demonstrate that PoS (either one) can give the same network safety guarantees as PoW.
Check the "Simulation" part of the Snow White paper I linked above for instance[1]. You're right to ask that, because indeed most Proof of Stake algorithms either have weaker[2] or slower[1] safety guarantees.
[1]: https://eprint.iacr.org/2016/919.pdf#section.5
[2]: https://docs.zilliqa.com/whitepaper.pdf