By "Nothing at stake" you mean the attack described in this link
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2393940 [first google result]?
If so, the attack requires 51% of forging stake, which would pump the price to very high amounts. Moreover, the attacker is stuck with 51% of forging stake, which in NXT would [after bubble] mean burning tens of millions of dollars. That sounds like a much-at-stake attack to me.
Or do you mean some different N@S attack? If so please provide a link.
Ad. 1)
Right now when someone synchronizes, he downloads the chain with the most cumulative hit. [cumulative hit == most POW work put into it]
To be able to fake cumulative hit you'd need to have > 50% of forging stake at any point in time since the last hard fork. An attacker would be burning tens of millions of $.
Ad. 2)
You cannot create blocks into the future and present them as valid to unsuspecting client. A new client checks the timestamps, and if they're invalid rejects the blocks.
[Assuming I understood your description correctly. If not, please elaborate.]
Ad. 3)
How would you
exactly implement this system?
Depending on the variant PoW miners may have hands tied so strongly that they have no defense against 51% stake attacker, or the attacker could attack the network with 10% stake & 51% PoW.
"The miners can't censor or be coerced to." - If miners can select which blocks get into the blockchain, they can censor. What prevents the government from building ASIC's? What prevents the governments from giving 51% of miners $ for their efforts?
1) Even if PoW would add security [it woud decrease it if implemented incorrectly], it'd rise the costs of keeping the currency considerably. Compare inflation of PoW currencies [8.2% currently] with for example NXT [0% inflation, financed 100% from fees] and you'll see why it'd be a system with considerable drawbacks.