I do think it would be constructive for dev teams to take these situations into consideration and have a plan and/or stated policy in place for how to cope with "existential threat" hacks like this.
First, there is no existential threat. At least not to Bitcoin. If Ethereum ends up voting to NOT invalidate the event (remember - it was clearly stated that the only binding terms were the code, which makes the 'heist' actually a legitimate action), then it may yet recover. If they do invalidate the recent event, nobody will ever trust again that it is decentralized.
More importantly, Bitcoin already has a policy. You make a stupid investment, it's on you. Don't expect the system to bail out your ignorant ass.
I didn't say it was an existential thread to bitcoin. That's silly. It was an existential threat to the entity in question, The DAO. Given that the head of Slock.it is now saying it's dead regardless of what happens, I'd say that's pretty existential.
If they engage a hard fork it will require widespread consensus among the miners involved to pull it off. There is no contradiction between decentralization and getting something done. I think the muddled point you are trying to make is that if they bail out The DAO investors, people will expect it again in the future. That risk exists both in a mob rule and a dictatorship. That, not decentralization or centralization, is the real issue at hand.
As long as they can do it with a soft fork it will be OK.
Same thing for deleting satoshi's coins, if theymos and mining pools can do it with a soft fork (i.e. not changing the codes of Bitcoin, which is immutable.), than is OK.
1 - I don't think it is possible to prevent certain coins from being spent, at least on a permanent basis with a soft fork (based on my limited knowledge about cryptography). My understanding is that, in order to prevent certain coins from being spent via a soft fork, at least 50.00...01% of the miners (or whatever higher agreed upon threshold for the soft fork to activate) need to agree to not include transactions from the blacklisted address(es) in their solved blocks, and to not build on top of any blocks that include transactions from the blacklisted address(es). However, it is my understanding that miners controlling 50.00...01% (or possibly less if they have a little bit of short term good luck) could decide to include transactions from the blacklisted address(es), and to build on top of such blocks at a faster rate then that of the competing blockchain, forcing the other miners to decide between building on top of the blockchain the moved the blacklisted coins and risking block rewards of found blocks that will never make it into the "final" blockchain.
On the other hand, I understand that a hardfork could make transactions that spend coins from blacklisted address(es) invalid so regardless of the length of the chain the builds on top of a block that includes transactions from the blacklisted address(es), that chain will be invalid, and miners will have clear incentives to build on top of the chain without these transactions because the economy will reject the other chain.
2 - The code of Bitcoin can (and has been) changed via a hardfork. The code of Bitcoin is simply code that has been accepted by the Bitcoin ecosystem. If there is a change to the Bitcoin code that the overwhelming majority of the Bitcoin ecosystem (and more importantly the bitcoin economy) agrees would be beneficial to Bitcoin (and bitcoin - lower case 'b'), then the code will be changed.
I think there is a fairly substantial difference between blacklisting the stolen DAO coins and blacklisting satoshi's coins.
In the case of the DAO, it is fairly easy to
somewhat cryptographically prove that funds were stolen by looking at the blockchain and the DAO smart contract. I also understand that it is possible to validate the conditions of the smart contract by looking at both the smart contract and the address the funds were sent to (eg it is not possible to create a smart contract that has an arbitrary deposit address).
On the other hand, it is impossible to tell if satoshi was actually the person who is attempting to spend his coins. It is the bitcoin private keys that allow satoshi to spend his bitcoin, and not any other verification method. It is far worse to intentionally deprive someone of their property for no reason other then that their property
might get stolen in the future then to deprive someone of property that fairly clearly does not belong to them.
It would also be bad to blacklist the stolen bitstamp coins (or the stolen bitfinex coins, or other coins that are claimed to be stolen) because in order to say that certain coins are stolen, you will need to believe the person who claims to be the "true" owner of said coins, and there is no real way of verifying that the "true" owner of the coins did not receive some kind of consideration in exchange for sending their coins to the "thief".
ETH is still very new, and blacklisting ETH coins would do nowhere near the damage as blacklisting BTC coins would do.