Some follow on from
The assumption in that work was that the common case in running behind a load-balancer is that a given client would continue to be routed back to the same avatica server instance. Sadly, this is not necessarily reality.
If the only load balancer technology available is only capable of an round-robin algorithm (or similar), we need to provide the information for a client to make a decision to return to the same server upon subsequent requests (e.g. fetching the next page of results).
Thinking more generally, the server which processed a given request is just general metadata. We could include things like the Avatica version, the "real" JDBC version information, etc.