Wire fix/e2e-tests changes into the all-flows CI runner

[ElioNeto]

Issue imported from tinyhumansai/openhuman#1865
Created at: unknown

---

Summary

app/scripts/e2e-run-all-flows.sh runs each spec in a separate Appium session (Appium restarts per spec). With the single-Appium-session reset model introduced in tinyhumansai#1859, the all-flows runner is now duplicating session setup ~17 times per CI run. It should switch to one session for the whole suite, matching the design called out in app/test/wdio.conf.ts.

Problem

Current e2e-run-all-flows.sh:

run "test/e2e/specs/login-flow.spec.ts" "login"
run "test/e2e/specs/auth-access-control.spec.ts" "auth"
run "test/e2e/specs/telegram-flow.spec.ts" "telegram"
# … 17 specs, each via e2e-run-spec.sh → spins up Appium, mock server, CEF binary

Each run invocation calls e2e-run-spec.sh which calls e2e-run-session.sh which:

1. Cleans a temp workspace.
2. Launches CEF.
3. Waits for CDP.
4. Starts Appium.
5. Runs WDIO on the one spec.
6. Tears it all down.

That's ~30-60s of harness setup per spec × 17 specs = ~10 minutes of pure overhead per CI run.

Now that resetApp(<userId>) produces a fresh-install baseline in-place, all specs can share one session — just run wdio once over the whole specs/**/*.spec.ts glob (which wdio.conf.ts already targets).

Solution

1. Replace e2e-run-all-flows.sh with a single-session invocation: build once, launch CEF once, start Appium once, run wdio once over the full spec glob.
2. Ensure spec independence — each spec must call resetApp(...) in before() with a unique userId. Audit + add as part of the spec-conversion sweep (Convert RPC-heavy E2E specs to drive the UI past login tinyhumansai/openhuman#1860).
3. Per-spec failure isolation — WDIO's mocha.bail = 0 is already set; one failing spec must not stop the suite. Verify via a deliberately-failing canary spec.
4. Artifact collection — e2e/helpers/artifacts.ts already captures per-test failures; confirm they don't collide when many specs run back to back.
5. CI workflow — update whichever GitHub Actions job invokes the runner so it uses the new single-session script.

Acceptance criteria

• Single-session runner — e2e-run-all-flows.sh (or a new e2e-run-suite.sh) runs all specs in one Appium session.
• Suite runtime under 10min on CI — measured before/after; expect a 5-10× speedup for the harness-setup portion.
• Spec independence enforced — resetApp(...) in every spec's before(); covered by a lint or grep CI check.
• Per-spec failures are isolated — failing canary spec exits non-zero without killing later specs in the same session.
• Artifact directories don't collide — each failure screenshot/source XML lives under its own per-run, per-test path.
• Diff coverage ≥ 80% — shell script changes, exercised by running the runner in CI.

Related

• Blocked-by (loosely): Convert RPC-heavy E2E specs to drive the UI past login tinyhumansai/openhuman#1860 (specs need to be spec-independent before the single-session runner can fan them in).
• Unblocked-by: test(e2e): make the harness actually exercise the UI past login tinyhumansai/openhuman#1859.