FoundationDB adheres to the highest level of consistency (linearizability) and, unlike Spanner or its derivatives, does not rely on clocks to achieve this.
I'm surprised FoundationDB is not mentioned in the article. Maybe it didn't help the author make his point, but it is both the most mature of the new breed of ACID+noSQL databases and completely free/open source.
According to their docs, FoundationDB only provides serializable isolation, so it isn't the same as the others in the post which offer strict serializable isolation (multi-key version of linearizability). Without strong clocks you can't have strict serializablity and scale beyond a single log (but one log can get you pretty damn far). TBH most people probably only need serializable transactions anyway.
I'm pretty sure FoundationDB still relies on the more traditional model of hardware: "many fast, identically-spec'd machines on a rack, with fast reliable networking".
In other words, I don't think you can run Foundation DB across geographically dispersed data centers, as you can with Spanner, CockroachDB, and others.
Running on cloud VMs is a very hostile environment for a database, as others have pointed out in this thread.
Though I'm happy to hear from someone who knows more about FoundationDB.
FoundationDB could probably do better than CockroachDB in multi datacenter configuration, because of the different trade offs they made, but I still wouldn't expect any of them do that well enough for most applications.
I'm surprised FoundationDB is not mentioned in the article. Maybe it didn't help the author make his point, but it is both the most mature of the new breed of ACID+noSQL databases and completely free/open source.