Sorry for misreading your questions. The article talks about blockchains in general, it does not refer to one in particular, so it is not possible to give specific answers to your questions.

When people talk about "general purpose" blockchains, usually they have in mind Ethereum (which uses cpu-resistant POW, to make sure ASICs are not economical. Planning to transition to POS) and Hyperledger which uses PBFT. My understanding of Hyperledger is that is assumes that most nodes are acting in good faith, but I might be wrong.

It's a VM, but it's not free. You need to pay the cost of CPU processing, so to be able to be 'unkillable', a virus will need to be able to generate revenues on its own, to be able to support its "life-style".

But yeah, being turing-complete is a major challenge. Ethereum sustained a major DOS attack a few months ago because some opcodes were mispriced. An attacker was able to bog down the network with a relatively small investment.

Yup that's the biggest hurdle right now but multiple ideas are being explored. The one that I personally find the most promising is "sharding", an idea developed to allow the Ethereum blockchain to scale massively. In a nutshell, they will split the blockchain in multiple shards and allow transactions between them.

More details here: https://github.com/ethereum/wi...

Woah that's a lot of questions. Let me try to answer the first two:

What is the "proof of work" used by the blockchain to decide which node gets to commit to the permanent blockchain record?
A complex mathematical question that takes a long time to answer. Complexity of the question is adjusted dynamically to aim for a regular block "commit" like you say. There are alternatives to PoW, like Proof Of Stake (PoS) where you "stake" an amount of money that you will lose if you cheat.

How will the blockchain handle if a pool of nodes consisting of > 50% of the computing power for proof of work decides to "double spend" or alter a blockchain record?
It happened on smaller, less popular blockchains, where the investment to hire the necessary computer power was relatively small. That's why bigger, more popular blockchains are considered more "secure". For bitcoin, that would cost hundreds of millions and you might gain at most thousands of dollars. Double-spending will be discovered a few blocks later, that don't leave you a lot of time to cash out the bitcoins.

Actually, it is quite the opposite. Blockchain platforms, like Ethereum, are being built to address the re-centralization problems that you describe.

Ethereum aims to build a centralized "world-computer" with 3 main components:
- ethereum blockchain: decentralized turing-complete VM, with built-in ability to create cryptocurrencies
- swarm: decentralized storage, similar to IPFS
- whisper: decentralized messaging

The project is still in its infancy, but it's already usable. Still not 100% decentralized as it still relies on HTTP and DNS, and scaling is on the roadmap. Just to give you an idea of what this platforms makes possible, here are examples of projects already launched or in alpha:
- OasisDEX: A decentralized cryptoasset exchange
- ENS: Decentralized Naming Service, to replace DNS
- uPort: self-sovereign identity
- Everex: remittance for migrant workers