1. Every full node is part of this dark forest, maintains its own mempool, and gossips the mempool to others. You do not have to be a validator or miner to run a full node; just run an Ethereum client.

2. There are indirect economic incentives to 'watch the watchmen': a slashable staking violation that doesn't get slashed breaks the Ethereum security model, and is expected to be an extremely rare event (if it ever happens). With a network that has a market cap in the hundreds of billions; someone is watching and it only takes one to sound the horn.

I do agree that there probably will be altruistic validators in some cases, but the economic incentives work against them. The history of cryptocurrency has trended towards doing the minimum amount of validation work that satisfies network consensus[0]. It's entirely possible that the market just shrugs its shoulders and says that a violation of the Ethereum security model just isn't actually that bad. Consider the relative lack of concern over MEV[1] and automated arbitrage bots, and the fact that the valuation of the coin is entirely unconnected to any security issues it may or may not have. It may just wind up being the case that the network just... tolerates a handful of people getting screwed.

[0] For example, Bitcoin miners accept new blocks by literally joining other mining pools purely to steal their previous block hash. It's significantly faster than waiting for a full 4MB block to get P2P gossiped around the world, but the downside is that you can't actually validate any of the data in that block. This effectively means that miners aren't full nodes anymore, and the result has been several long chainsplits and reorgs whenever a softfork occurs.

[1] Miner Extractable Value

Following this logic, stakers themselves are incentivized to just run the software to submit a slashing tx themselves (free money), at which point just having stakers running slashing software is sufficient. And if they aren't someone else will submit it to the mempool.

Does it matter who does the slashing, as long as the incentive is enough to motivate doing it?

Ok, let me try ELI5ing it.

Let's say we have an economy where, in order to determine who owns what, people literally just make and publish a list of who owns what. Everyone trusts the lists. But what if someone decide to change the list so that it says they own someone else's stuff? Proof of Stake says that if someone wants to change the list - say, if someone traded something on the list to someone else - the page they add to the list also has to have some of their money taped to it. If that change turned out to be fraudulent, someone else can tattle on them, and then that someone gets the money taped to the list instead. So if you want to make sure you're reading from a not-fraudulent list, just check for the money taped to the page.

This all continues fine for a while. There's a bunch of people who maintain the list, all of whom have their own money taped to different pages of it.

One day, someone tries adding a fake transaction to the list. You, of course, notice it and tattle on them. Except in order to claim the funds taped to the list, someone else has to... add a note to the list saying you own them. And there's no reason why they can't just write down that they tattled instead of you, because the only proof that you were the one who tattled is getting it onto the list that you have to tape money to in order to add to. So the only people who have a reason to make sure the list is valid are the same people who are updating the list in the first place. Everyone else can tattle but they don't get paid for doing so, which was the the thing keeping the system secure.

As an interesting aside, this is essentially how the economy of the Yap islands in Micronesia worked until the 20th century, with large stones from another island changing 'ownership', while being left in place (because it was too hard to move them)[1].

Except it was an oral history instead of a published list!

[1] https://en.wikipedia.org/wiki/Rai_stones

It reminds me of the technique to remembering a book where you read one word at a time and repeat it out loud, repeating each segment + 1 in a loop. It’s quite trivial to remember every single word when doing that.