I just realised that my point 2 has already been considered of in the PDF too:
One strategy to protect against this would be to accept alerts from network nodes when they detect an invalid
block, prompting the user's software to download the full block and alerted transactions to confirm the inconsistency.
Another question, which I had forgot to ask: Could attacker nodes create new coins to spend? If a huge amount of CPU power was maintained, presumably the bad nodes could validate invalid Bitcoins (their own, cheaply minted)? Or would this be impossible, as even a thin client (the 'Simplified Payment Verification' clients, which haven't been programmed yet) would flag such coins as invalid?
Thanks again!
Anyone have a good answer for the above? It would complete my knowledge on how it all hangs together and it would give me confidence in the system.