An intentionally dishonest article.
The actual truth is that PoS is infinitely safer than PoW in the short to medium term, while theoretically weaker in the long term. A long-term attack would require first buying obsolete signing keys, which would stop nodes that sync starting from the pre-fork point from syncing - ie. a denial of service attack. Which is in a very weak threat, as online nodes wouldn't even notice it.
A short to medium term attack would stop finalization for a while at an enormous cost of slashing.
It's a denial of service attack because nodes would be able to see contradictory signing from the same keys - so while without out of band data they won't be able to decide which one is the commonly accepted chain, it's enough information to recognize than an attack is happening.
PoW is very weak in the short term to medium term because runtime cost of attack is equal to mining rewards + epsilon, which is negligible, meaning it's just a question of hardware. Contrary to PoS, mining hardware is an external resource - it's always possible to get enough of it, given enough money (single digit billions for bitcoin). Getting 2/3 stake of a long-running PoS system is impossible - it's a scarce internal resource and there isn't enough for sale.
Reverting years of blocks is indeed infeasible - but interestingly in practice it would also amount to a DoS attack, as everyone would notice it and pause all payments. Contrary to PoS, where it would only work on newly syncing nodes, it would stop everyone. However, while theoretically more expensive, it's still only a matter of money - while a long-run DoS attack against newly syncing nodes in PoS would require buying obsolete keys, which is very likely to be impossible in practice.
Is this even an advantage? I don't think so, but it's arguable. However, for this singular arguable point PoW pays with a 4 orders of magnitude higher cost and a much, much weaker short and medium term security.
Empirically, lower security of PoW is confirmed: multiple 51% attacks happened (most famously ETC), while even a much weaker DPoS coins never had a successful double spend attempt.
In terms of public trust, not many people are able or even interested in technical arguments - they just observe if something works. In reality, consensus-level attacks are very rare as it's currently very hard to profit from them regardless of the consensus method, and the biggest danger is from software bugs in nodes, most likely unrelated to consensus.
If any PoW blockchain became a foundation of global commerce, attacking it would become very profitable, or even a military target - but that's never going to happen. So I don't expect bitcoin to get 51% attacked in any near future - at best years in the future when value of block rewards is so low one person with lots of old mining hardware can attack it just for fun.