Yes, this is common. That's a throughput-latency graph. What isn't being plotted in that graph is that there's actually an underlying controlled variable which both throughput and latency are responding to---the number of clients/outstanding requests at any given time.
The reason you're seeing a "wall" is that the system has some peak throughput, and once you reach full saturation, the average latency for requests is going to go up to infinity. (So, it's actually not exponential growth, it's approaching an asymptote.)
The reason you're seeing a "wall" is that the system has some peak throughput, and once you reach full saturation, the average latency for requests is going to go up to infinity. (So, it's actually not exponential growth, it's approaching an asymptote.)