This probably belongs to development forum, but lots have changed here over the last year apparently...
Two questions about network logic.

Can anyone explain to me why this scenario cannot happen:
The hashes are practically random, for every block there is 32 bit of nonce and 32bit of timestamp.
Can it happen so that none of 18446744073709551616 hashes that would be generateable from this set of variables would be below the target?
It sounds like a freak of luck for that to happen, is this considered or is deemed too unlikely?


Second, in practical terms, does this put a limit on generation rate at roughly 4.2 Ghash/s?
Once all the nonce is tried, the timestamp must be incremented, but a second is not done yet.

With network rate of ~10Thash/s, does that mean that there is a lot of redundant tries?
With deviation in time spread about +-5 minutes that's still 600 possible sets of 4.2 Ghash/sec domains, naturally with lots of repeats.
What have i missed?