I see the ideological foundation for a theoretical 51% attack more aligned with causing mass disruption & chaos than some kind of financial gain. Yes, it would likely be detected. But it would undermine BTC, lots of people would lose lots of money, lots of businesses would go under, portions of the global banking system would be disrupted and likely pushed back to more centralized alternatives. The most likely threat actor for such an attack would therefore be a nation-state.

China is the only country pushing that kind of hash rate at the moment, so we should analyze whether the move would be beneficial to them in order to determine the risk.

You are assuming that the attacker would not gain more from a general loss of faith in Bitcoin than they would gain from controlling Bitcoin itself. That is a bad assumption. What if the attacker is trying to push a competing system? What if the attacker holds short Bitcoin futures contracts?

> But a 51% attack (with e.g. a double spend) will be easily spotted by the community, right?

The real problem is that if it happens once, it's proven to be both possible and practical. And once it's proven to be possible and practical, how do you know it won't happen again and again?

> What would be the incentive for an actor with such a vast amount of specialised hardware just for Bitcoin mining, to undermine its security?

Most mining hardware is specific to hashing, not Bitcoin specifically.

The obvious attack is something like this: Imagine someone invents a Bitcoin competitor that is somehow more resistant to these types of attacks, yet can use the same mining hardware. To convince everyone to switch to their new alternative (which they have accumulated significant amounts of, similar to Satoshi), they spend the money to crash Bitcoin, while advertising themselves as the more secure alternative to Bitcoin.