there are certain timestamp validity checks:
https://en.bitcoin.it/wiki/Block_timestamp

Difficulty is based on the number of blocks created, and is adjusted every 2016 blocks.

Thanks for the correction. However, I'm going to add a trivial correction to your correction   Difficulty is based on how long it takes to generate the 2016 blocks. If it takes longer than expected (two weeks), the difficulty is reduced. If it takes less than two weeks, then the difficulty is increased.

In the end I'm not saying that would be  a huge problem in that it would basically just jump us ahead 2 weeks but how possible is that for someone to pull off? It seems like a pretty good sized vulnerability. Especially if a couple large miners all decided to do it.

Good, it seems like everytime I get a little worried there's a fix built in to the system.

Yeah it still worries me a little because I feel like it's impossible to have thought of everything.

Doesn't seem worth it to me like that.

Even if this were a realistic exploit, the difficulty has an adjustment rule that prohibits an adjustment of greater than a factor of four in either direction.  So the difficulty couldn't be cut down to less than a quarter of the prior difficulty.  It's still something to consider, as this is one attack avenue I don't think that I've seen discussed on this forum before.  But before we consider how difficult the attack vector may be, we need to determine what the goal would be.  If the goal is simply to spoof the difficulty calculations in order to make it easier to capture some blocks in a short period of time, (perhaps someone decided to buy a couple of hours of mining capacity on Amazon's cloud?) then really no great harm done, and if such things were to become an ongoing thing, a tighter difficulty adjustment parameter could undercut such activities.  However, if the goal was to actually harm the network in some fashion, the impracticality of the attack may not be a great defense.  How might this harm the network. or introduce another fraud vector?  thoughts?

Are you implying that the 'Mystery Miner' was such an exploit? 

Okay, so how would this attack vector aid in a double spend fraud? 


How?

Except that can't work.  As I mentioned earlier, there is a difficulty adjustment parameter rule that prohibits the difficulty from adjusting up or down by more than a factor of four.  So the max that can be expected is that the Google can do is move the difficulty by that factor of four, which may or may not actually be worthwhile, but if the attack cannot be repeated in consecutive cycles (maybe, but I would say that it would be very unlikely to work out that way) then it's probably not a worthwhile means of manipulation for profit motives alone.  What kind of harm to the system itself could such an attack cause?