test(gateway): empty turns are classified failed since #270

[OmarB97]

What does this PR do?

Fixes test_prompt_submit_preserves_empty_response_without_error, which fails
on main with assert 'error' == 'complete'.

This is a stale expectation, not a regression — and #270 says so in its own
tests, which is how I confirmed it rather than guessing.

#270 ("guarantee terminal turn outcomes") made the turn-outcome classifier the
owner of message.complete's status. _derive_turn_outcome() classifies a
turn that ends with no visible response as failed:

else:
    status = "failed"
    reason = fallback_notice or error or raw or "turn ended without a visible response"

and _freeze_turn_outcome() then overrides the locally computed status
(tui_gateway/server.py:10930), mapping failed onto status: "error". So the
old result.get("error")-based computation no longer decides the outcome.

That this is deliberate is settled by #270's own test file:

# tests/tui_gateway/test_turn_outcomes.py
(
    {"completed": True, "final_response": "", "messages": [...]},
    "failed",
    "without a visible response",
),

#270 added tests/tui_gateway/test_turn_outcomes.py (800 lines) but never
touched tests/test_tui_gateway_server.py, so this pre-#270 expectation stayed
red — invisible in CI because the uv sync --locked install step had been
failing since 2026-07-10 (see #286).

The half of this test that still guards real behavior is the payload text:
classifying a turn as failed must not fabricate an "Error: ..." string. That
assertion is unchanged and still passes.

Related Issue

No filed issue — surfaced once #286 unblocked the Python test slices.

Type of Change

• ✅ Tests (adding or improving test coverage)

Changes Made

• tests/test_tui_gateway_server.py — expect status == "error" for a turn with no visible response, and document in the docstring that fix(runtime): guarantee terminal turn outcomes #270's classifier owns this, with a pointer to the authoritative parametrized case.

How to Test

pytest tests/test_tui_gateway_server.py -p no:randomly -q -k test_prompt_submit_preserves_empty_response_without_error

Before: assert 'error' == 'complete'. After: 1 passed.

⚠️ Run with an isolated HERMES_HOME. The suite reads the developer's real
~/.hermes/config.yaml; a local fallback_policy: 'off' reds 5 unrelated
tests in this same file that pass on CI.

Checklist

Code

• I've read the Contributing Guide
• My commit messages follow Conventional Commits
• I searched for existing PRs to make sure this isn't a duplicate
• My PR contains only changes related to this fix
• I've run the affected test and it passes
• I've added tests for my changes — N/A, this PR repairs an existing test
• I've tested on my platform: macOS 15 (Darwin 25.6.0), Python 3.11.15

Documentation & Housekeeping

• I've updated relevant documentation — N/A
• I've updated cli-config.yaml.example — N/A
• I've updated CONTRIBUTING.md or AGENTS.md — N/A
• I've considered cross-platform impact — N/A, test-only
• I've updated tool descriptions/schemas — N/A