Vespa's Document API provides single-document operations, in other words if you want to index N documents you have to issue N http requests to it.
Does the Vespa feed client work in the same way? I.e. does it issue N requests, but does so efficiently over HTTP2 / etc?
Yes, you are right - operations are individual, there is no batch feed api. HTTP2 is recommended for efficient feeding, as it can handle multiple simultaneous streams per connection.