Sunny, theoretically, won't the primes get bigger as we find more of them, so that it's impossible to maintain a uniform time in between blocks? So that in fact, if the network hash rate remains constant over some period of time, difficulty will adjust lower?
Difficulty is measured by chain length, not by prime sizes. Block spacing is maintained via continuous difficulty adjustment.
Sunny, I keep observing periods where the block spacing is ~30seconds instead of ~60seconds, yet the difficulty isn't increasing by a corresponding factor.
Example period where block spacing is ~30seconds:
block 665 creation time = July 8, 2013, 04:11:39 difficulty = 7.03636408
block 675 creation time = July 8, 2013, 04:17:56 difficulty = 7.03710973
It has been a long time since blocks were spaced according to your specification so it would appear the variable difficulty should be set higher.