Well sure. Consistency is great and for years most DBs people really used gave you all of the ACID guarantees. Then, huge scar came along. At extreme scale, sometimes you have to trade things for speed. But that extreme scale is a problem that frankly most people don't have.
So don't trade off a good attribute unless it's absolutely necessary. Seems sensible
It's really hard to convince people they don't have (and aren't likely to have) a Facebook-sized scale problem though, even if the vast majority of people really don't. It's like trying to convince people they don't have "big data". :-)
This bridges over to another thread (https://news.ycombinator.com/item?id=16156972) about the problem with many IT people thinking of their job as playing with cool technology rather than business outcomes. I’ve run into even a fair number of high-ranked people who habitually over-engineer things because they’ve tied their ego to the wrong metric, and many places didn’t have corrective pressure against that.
Hosted services have been good for counter-pressure: if your oracle cluster can be replaced with a $50/mo RDS instance, it’s a lot harder to hand-wave around the cost differential.
So don't trade off a good attribute unless it's absolutely necessary. Seems sensible