The streamed report needed no change, and now there is a test saying so

Two behavioural changes upstream: report streams its full record set over the socket rather than a summary, and a serving build is interactive-first - it listens immediately and runs the suite only when asked.

Neither needed a code change. Records between ack and done were always collected, and orbistoun session asks for the report rather than expecting one on connect. But "no change needed" is a claim, and an untested claim about a path that has just become the primary one is worth converting into a test.

a_streamed_report_arrives_between_the_acknowledgement_and_the_answer drives a full stream

• section, try, res, sym, sectiontally, tally - through the client, replays the wire into the same reader a committed corpus goes through, and asserts sections, findings and symbols all come out. It also asserts that tally, a kind this version does not model, survives verbatim; that is the property letting obSCEne add records without breaking this.

nothing_is_expected_to_arrive_before_a_command_asks_for_it pins the other half: a client assuming a report on connect would block forever against a probe behaving correctly.

Surprises

The valuable part of "no change needed" was writing the test anyway. Both changes landed on paths that already worked, so the honest options were to say so and move on, or to make the claim checkable. The second cost fifteen minutes and turned an assumption about the newly-primary path into evidence - and this session has spent most of its time discovering that untested claims were wrong.