Could you tell me more about the rationale of keeping unknown config fields?

Could you tell me more about the rationale of keeping unknown config fields?

Two reasons: (1) as you can see from the removal of this code:

    # These keys are top level in job resource and query resource.
    if "labels" in job_config_resource:
        request_body["labels"] = job_config_resource["labels"]
    if "dryRun" in job_config_resource:
        request_body["dryRun"] = job_config_resource["dryRun"]

it allows for a cleaner conversion from query job insert request to jobs.query request.

(2) It allows for customers who want access to features that are in preview an not yet implemented in the client library to set REST API properties via setting QueryJobConfig._properties. This is a very common practice (see: #1398 (comment) for example) and is in fact a requirement for features that are in a "private" release. We don't support adding private features to our public client library, but this _properties feature allows such customers to use the public client library via this workaround.

That's great to know! Thank you