It still works like this and they have plenty of headroom with the new solution. Chances are good that when they need to upgrade this solution, technology will

It still works like this and they have plenty of headroom with the new solution. Chances are good that when they need to upgrade this solution, technology will have advanced far enough as well. Guessing what their performance demands are (far, far more reads than writes) this seems to work fine for them, so why make it more complicated? They do talk about read-only replicas, so they have some distribution.