Testing plan: the hosted ingress set follows the config

Derived from bugfix.md and design.md. Planned at test-planning, results recorded at verification (below).

Test matrix

#	Type	Applies?	Scope / what it proves	Where it runs
T1	Unit	yes	`_wanted` composes the same list for boot and reconcile; `HostedIngresses.reconcile` stops what is no longer wanted and starts what is newly wanted, drops a dead entry, and leaves a same-membership change alone	`cd cli && uv run pytest tests/test_hosted_reconcile.py`
T2	Integration (scenario, Gherkin-docstringed)	yes	a real config file, a supervisor with a short interval, the file rewritten: `read.mode: off` ends the listener's loop and releases its lock with no restart; `socket` again starts it with the new config; an unrelated edit and an unparseable edit change nothing; `polling.enabled: false` stops a hosted poller; shutdown stops the supervisor then the ingresses (R1.1–R1.6, R1.8)	`cd cli && uv run pytest tests/test_hosted_reconcile.py`
T3	Contract (OpenAPI)	n/a — no API route or schema change; `status --format json` is unchanged (R2.2)
T4	End-to-end (live)	no — the Socket Mode client is faked at the seam issue-334's tests fake; the live B5 reproduction is the next Slack run's
T5	UI / visual	n/a
T6	Snapshot	n/a — the `status` line is asserted directly (T7)
T7	CLI rendering	yes	`the-loop status` prints the non-contradictory flag for a running-but-disabled row, hosted and standalone, and the plain flags otherwise (R2.1)	`cd cli && uv run pytest tests/test_lifecycle_cmd.py`
T8	Security / abuse case	yes	an unparseable edit changes nothing (abuse 2); a reconcile whose starter raises records `ingress.hosted_failed` and the supervisor survives (abuse 3); the missing-token refusal on a reconcile start names the variable, not the value	with T2
T9	Accessibility	n/a
T10	Migration / upgrade	n/a — no key, state or schema change; `build_lifespan`'s new argument has a default
T11	Manual exploratory	no — deferred to the next live Slack run (B5's own steps)
T12	Regression (full suite)	yes	the existing hosted-listener, hosted-ingress, lifecycle and SDK suites stay green: `start_hosted_ingresses` / `stop_hosted_ingresses` keep their contract	`cd cli && uv run pytest -q`

Scenarios & requirement trace

Row	Requirement(s)	Scenario / case
T1	R1.2, R1.3	`_wanted` for {nothing, poller, receiver, socket listener, poll-mode listener} equals what `start_hosted_ingresses` composes
T2	R1.1, R1.4, R1.8	`Scenario: read.mode goes from socket to off` — the listener's stop event is set, its thread ends, its lock is released, `ingress.hosted_stopped reason=config` and `config.reloaded` are emitted, the service process is still up (no restart)
T2	R1.2, R1.4	`Scenario: read.mode comes back to socket` — a second listener starts, under the same lock, with the edited config; `ingress.hosted` emitted
T2	R1.3	`Scenario: an unrelated key changes` — same thread object still alive; no ingress event
T2	R1.5	`Scenario: the edit is not valid YAML` — nothing changes; the next valid edit is applied
T2	R1.1	`Scenario: polling.enabled goes false` — the hosted poller's stop event is set and its lock released
T2	R1.6	`Scenario: shutdown` — `stop()` ends the supervisor thread before the ingresses; no ingress is started by a tick after `stop()` returns
T2	R1.2, T8	`Scenario: the newly wanted listener has no tokens` — `ingress.hosted_failed` names the variable; the supervisor keeps ticking
T7	R2.1	hosted + running + disabled → `disabled in config — still running; the service stops it on its next config check`; standalone → `\`the-loop stop` ends it`; enabled rows unchanged
T12	R1.7, all	the whole suite

Verification environment

Repositories: this repo only.
Services / containers: none; the Socket Mode client and the poller's run loop are faked at their seams, and the supervisor runs in-process on tmp_path.
Fixtures & data: a cli-config.yaml written per test under tmp_path.
Credentials: by reference only — THE_LOOP_SLACK_BOT_TOKEN, THE_LOOP_SLACK_APP_TOKEN, set to placeholder strings by monkeypatch.
Bring-up: uv sync · Tear-down: none.
If bring-up fails: record under Verification results, leave the rows unticked.

Evidence plan

Row	Evidence	Path under `evidence/`
T1, T2, T7, T8	test names and run output, red → green noted	`automated-tests.md`
T12	full-suite, lint, format, typecheck output	`automated-tests.md`

Verification activities

[x] T1 + T2 + T8 — cd cli && uv run pytest -q tests/test_hosted_reconcile.py
[x] T7 — cd cli && uv run pytest -q tests/test_lifecycle_cmd.py
[x] T12 — cd cli && uv run pytest -q · uv run ruff check cli hooks · uv run ruff format --check cli hooks · uv run pyright cli

Verification results

Recorded in evidence/automated-tests.md.

Activity	Command / procedure	Outcome	Evidence
T1 + T2 + T8	`uv run --project cli python -m pytest -q cli/tests/test_hosted_reconcile.py cli/tests/test_hosted_listener.py`	pass — 14 passed (8 new, red before the fix)	`evidence/automated-tests.md`
T7	`cd cli && uv run python -m pytest -q tests/test_lifecycle_cmd.py tests/test_instance.py`	pass — 72 passed	`evidence/automated-tests.md`
T12	`cd cli && uv run python -m pytest -q` · `uv run ruff check cli hooks` · `uv run ruff format --check cli hooks` · `uv run pyright cli`	pass — 4291 passed, 1 skipped; ruff and pyright clean; CI green on the PR head	`evidence/automated-tests.md`

Review comments

None yet.