Thanks for the vote of confidence guys, but I'm having trouble following the question and answers. Can someone boil post a precis of the question?
In short:
I noted that we had a 35 second block and suggested that this might be some sort of record - another poster mentioned that we once had a 2 second block, but it was invalid.
This got me thinking that this situation must be extremely unlikely to happen, since the odds of a block lasting 2 seconds are quite short, and the odds of two pools finding a two second block at the same time would be even shorter. A quick estimate of the odds, if my understanding of the maths is right, is that the probability of a block being 2 seconds or less is around 0.001, so two at the same time would be around 0.000001, or 1 in 1 million (as pointed out by another poster, the odds of there being a 'race' between these two would be close to one because the block time is so short, rather than 1 in 60, which seems to be the approximate rate of orphaned blocks in the network)
This led me to wonder whether there is something about the (double) SHA256 algorithm that might lead to hash collisions being more likely for certain blocks, so that these blocks rather than being expected to last on average 10 minutes, would have much lower than expected difficulty. This might shorten the odds of there being an orphaned 2 second block. I don't know how possible this actually is, given that the block contains the address of the pool doing the hashing, so that the actual block being computed by two pools would differ.
So, I suppose what I am asking is whether my estimates seem fair, or if I have got the maths all wrong, and whether it is plausible that the hashing algorithm could lead to collisions in this way, causing the occasional block to be much easier to solve, for certain pools, or the network as a whole. I appreciate that cryptanalysis is far from a simple topic, so the second question is really more rhetorical.