Again, »power concentration« is not a binary thing. Is a network controlled by 300 independent actors better or worse than one controlled by 3000 people? What if the former requires an attacker to spend $300m and the latter just $30m? Or when the smaller network can come to consensus faster than the bigger one?
I’m definitely all in favour of making systems decentralized but everything comes at a cost and trade-offs need to be considered.
I think the biggest problem with the ASICs is exactly the issue of scale and hardware manufacturers having strong incentives to pre-mine with them, while also massively increasing the barrier to entry. Maybe if ASICs designs were released together with the PoW algorithm, then the asymmetries might not be as bad. But in the end, whoever can scale their operations the best will win the race for blocks.
To give you an example where ASICs might be better, take the Goldfinger attack, where an attacker—in our context here not being in possession of ASICs or a majority of the hashrate—wants to destroy confidence in the network. If we ignore the fact that a lot of crytpocurrencies use the same PoW algorithms, then their hardware would end up worthless after the attack. Therefore it would be more expensive than an attack against a network secured by general purpose hardware, given the attackers ability to easily rent such hardware or buy and later sell it again.
One of the big advantages of PoW for cryptocurrencies is that it makes bootstrapping, and therefore participation, easy, at least in principle. There is no need to acquire coins to take part in the consensus protocol, something you’d need to do for Proof of Stake. All you need is just a piece of hardware that can solve the puzzles.
Your last point seems off. Take facebook as a popular example. They have enabled all sorts of questionable behaviour via their platform, yet people keep using it, exactly because they can’t just go elsewhere. Their data is in the walled garden that is facebook.
Forking an open source system—although there are definitely important issues to be considered in the case of contentious forks—is entirely possible and has been done successfully. If you fork your cryptocurrency and change the PoW algorithm, because some powerful player with sufficient hashrate to harm the network starts misbehaving, then the majority of the userbase will most likely follow you and the malicious player would be left alone with a dead fork, that nobody trusts.
Some caveats apply here, but the overall sentiment should be clear.
In the zcash forums there was/is also a big discussion about this, since they recently found out that ASICs for their PoW algorithm exist.
There is a lot more to this discussion and I hope that I don’t come off as a big proponent for ASICs, which I am not. I’m just trying to make the point that things are complicated.