Confusion about architecture

Hello everyone,
I am now working on Thingworx architecture investigation, and I'm got a problem:
I use two connection server for device connectivity Load balance, I believe with this deploy mode, the architecture here is scalability

After that, the message received by connection server is send to Thingworx server. I know Thingworx server has a master-standby election based on zk-cluster

But in this way, I think there is only one live node(master) at a time, so I believe this is the performance bottleneck here, this can not expand dynamically

I don't know if I understand this right or not, can anyone help me to explain this

Thanks in advance