...
First, it would be pointless imo to argue against the facts: mining is concentrated more than most expected, say, 2 years ago, and more than many are fully comfortable with.

Fact as well, however: none of those conjectured 'mining cartel attacks' ever took place, and there's good reason for that: they might be technically possible, but economically not viable for anyone who holds a stake in Bitcoin (hint: sunk cost in single purpose hardware is a "stake" as well) because of the resulting market reaction upon discovery (or even suspicion such attacks are taking place). They would therefore only be mounted from someone who is willing to pay for an attack on the network itself, and I have yet to see any evidence that this is a realistic vector consistent with what we know about existing hashpower and who paid to employ it.

Finally: the level of decentralization doesn't look so bad anymore when one considers the existing alternatives. Well, there's only one really: traditional banking. It's like you're blaming jet plane air-travel for being hopelessly outdated and slow: until there's a superior alternative, it is the de facto optimum wrt speed.