Can I extend an already created cluster?

Extending a cluster is not supported.

What happens if a non-primary platform VM becomes unavailable?

Internal services have limited resiliency to non-primary node failures. In general, due to a node failure NI loses compute power.

What kind of load balancing is supported?

Mapping of proxy to platform is fixed. Once data from any Proxy VM reaches any Platform VM, its processing is load-balanced internally among all the Platform VMs.

Will creating a platform cluster increase bandwidth consumption?

The proxy or collector VMs continue to talk only to primary or platform VM. The bandwidth requirement for platform VM clustering communication is minimal. Therefore there is no significant increase in bandwidth consumption.

What is the frequency of data transmission between proxy VM to platform VM?

The proxy VM sends the deduplicated or compressed data continuously to the platform VM.

Does any optimization of data take place in the proxy VM?

Various deduplications, compressions, reduction, or batching steps happen in the proxy VM. When the connection between the platform VM and the proxy VM is down, the proxy VM stores data locally (depending on the disk space) and sends it whenever the connection is restored.

Is there any optimization done for network bandwidth?

Yes, various dedup/compressions/reduction/batching steps happen on Proxy VM.

When deploying a cluster , how do vCenter sends traffic to the various Proxy?

Actually proxies get connected to vCenter to fetch the information. Respective proxy will connect to designated vCenter & fetch the information. There is no clustering available on proxy