By database layer, do you mean the RDS in the diagrams? If so, they were logical diagrams; the deployment itself was more complicated to handle the realities o

By database layer, do you mean the RDS in the diagrams? If so, they were logical diagrams; the deployment itself was more complicated to handle the realities of AWS and whatnot. Still, having a single beefy RDS instance is a pretty common pattern for apps at this size. I've never experienced RDS postgres as a bottleneck for standard microservices architectures even at the 100-million-MAU scale.