The time to produce the next block is completely unaffected by the number of unconfirmed transactions.
The time is generally expected to follow a Poisson distribution as explained in answers to a related question.
I’m no statistician so what follows may be nonsense:
A Poisson distribution with a lambda of 6 shown as a probability mass looks like this
So in a thousand hours I might expect to see
- 160 hours where 6 blocks were produced
- 45 hours where 2 blocks were produced
Either that or something odd has happened, like a couple of large mining pools going offline.
It would be interesting to gather some of these sort of stats from the blockchain record but I expect this has already been done somewhere.