I'd be biased to agree since that's Manuel's blog, someone I used to work with. I've supported 24x7 and 9x5 ops where downtime was unacceptable. zfs makes it a whole lot easier to perform upgrades, know data and metadata are solid and send snapshots around.
Yeah, it depends on the use case. For home directories and large risk items like financial stuff, testing that barrier writes are happening is a good thing. You don't want a storm to knock out a DC to learn that the hw/sw fs stack was lying to you at some level.