Nope. That's very inaccurate. If only the ETH foundation is allowed to slash, then it's just a centralized system with extra steps. Because only the ETH foundation is allowed to decide what is bad behavior.

It is not the case that only the ETH foundation can slash, which was the crux of my original question.

The fundamental misunderstanding people seem to be having is that slashers show objective proof of bad events. That's accurate. But validators are still free to agree or disagree with that objective proof of bad events. If you can create a means to prevent consensus on slashing for your overtly malicious behavior, your stake is only at risk if the entire community agrees to fork, which is probable for huge malicious behaviors but I'm not convinced is true for smaller things or circumstances where large actors have a vested interest in preferring the original chain.

I agree with your 3rd, 4th, and 5th sentences, but not with your 1st and 2nd. Assuming a large number of independent validators, the Nash equilibrium is close to "there is no equivocation, hence no slashable evidence, hence no incentive to run a slasher node". (Remember that Nash equilibrium implicitly assumes that agents are not allowed to cooperate with each other). Suppose some fraction of the validators are deviating from this equilibrium by doing lots of equivocation, now there is an incentive to run a slasher node. So what this modelling suggests is that in the real world, we can have a small number of slashers and a large number of validators, and the security comes from the fact that anyone could be a slasher (it's OK that the number of people who actually are is small). But we cannot conclude that "it's OK to have a small number of validators, as long as anyone can be a validator".

I don't really agree but sure, however this counterpoint retracts the original premise, that the slashers are centralized. Something can be done with a small number of people and still be decentralized.