I don't think distributed databases are aiming primarily at increasing write throughput over a single node. The question is, how many transactions can you write per second from your California datacenter to your New York postgres leader when someone typos the BGP rule controlling that database's IP address? Probably zero. That's the performance number these databases are trying to increase.

Many people can say, "wait, that doesn't matter to me at all, that happens at most once a year and all of my customers are also offline when it happens". Indeed, that's why people are doing just fine with postgres. But, some things are a little more mission critical than average, and so these technologies exist for them.

One thing that does tip the scale towards these distributed databases are operational concerns. "Apply security patch" or "migrate to new major version" look a lot to distributed systems like "tornado took out the data center" or "hard disk turned into many disconnected chunks of steel". So while you might not be super concerned about disasters, you might still be interested in keeping your compute nodes disposable or in doing regular maintenance without downtime. I don't think the operational balance is quite there yet, but it's definitely worth looking into every few months to see what the state of the art is.