[riwsky]

Because the incidence and cost of mistaken under-consistency are both generally higher than those of mistaken over-consistency—especially at the scale where people would need to rely on managed off-the-shelf services like aurora instead of being able to build their own.

[wowamit]

I would be hesitant to generalise that. There is an inherent tension with its impact on the larger availability of your system. We can't analyse the effect in isolation.

[roncesvalles]

Most systems can tolerate downtime but not data incorrectness. Also, eventual consistency is a bit of a misnomer because it implies that the only cost you’re paying is staleness. In reality these systems are “never consistent” because you often give up guarantees like full serializability making you susceptible to outright data corruption.