The difference above - frak, re-reading it I note that I failed to actually make it clear. The scenario above with the dual RAIDZ1's of three disks each means that if a disk fails, you then have a 2 in 5 chance that a second disk failing kills the pool. The striped mirror vdev scenario reduces that risk AND significantly improves performance, but offers less space.
Once we get up to RAIDZ2 the risk model is substantially mitigated because you still need two more failures at a minimum to lose. But let me answer the question I'd like you to think about a bit:
If you have 12 4TB drives in 2 x 6 x 4TB RAIDZ2, that gives you two 16TB vdevs or a 32TB pool with the ability to lose any two disks. You may be able to lose up to four disks as long as two are from each vdev.
I propose that an 11 device RAIDZ3 plus a warm spare may be a better alternative. It still gives you a 32TB pool but the ability to lose ANY three disks, PLUS the first time you lose a disk, there's a warm drive standing by for rebuild.
The downsides to RAIDZ3 are that RAIDZ3 performs more poorly than RAIDZ2, and definitely even moreso compared to a striped dual RAIDZ2 setup.
So the question needs to be, what's the goal? For archival storage, I really like the 12-drive RAIDZ3 (11 + spare). For performance, RAIDZ2 still isn't great and I'd probably rather go three way mirror. I'm not a real huge fan of the RAIDZ2 options as a result. If I had a 24 or 48 drive chassis and I needed the space and also better performance than Z3, then 6 disk Z2 vdevs would be definitely something to consider.