webdriverio@9.27.2, @wdio/cli, @wdio/local-runner, @wdio/mocha-framework,
@wdio/spec-reporter, @wdio/appium-service, appium@3.5.0, @types/mocha@10,
tsx@4.22.4. xcuitest driver 11.10.0 installed to ~/.appium.

Add WebdriverIO 9 + Appium harness under tests/e2e/ios-appium/:
- helpers/app.ts: udidForName, requireSmokeUser, iosCapabilities
- helpers/elements.ts: el/els/waitForTestID/countTestID/byText/tab
- helpers/navigate.ts: navigateTo{People,Chat,Files,Teams,More} using real
  testID values from test-ids.ts; escapeToTabs with pointer swipe
- wdio.conf.ts: wdio 9 config (local runner, mocha, spec reporter, appium svc)
- tsconfig.json: standalone config for wdio files (excluded from main tsconfigs)

Also adds ios-appium tsconfig to eslint parserOptions.project so lint
covers the new files. ios-appium dir excluded from tsconfig.desktop.json
and tsconfig.native.json to avoid missing-browser-global errors.

…kers, reset-to-root, auto APPIUM_HOME

…abs before each spec, auto APPIUM_HOME

…b; native scroll helper

…a horizontal ScrollView

Each Kb.Tabs tab forced width:100%, which shrinks-to-share in a bounded row but
in the team/channel mobile horizontal ScrollView left every tab at full screen
width, pushing all but the first off-screen (the Settings gear was unreachable).
flexGrow:1 fills the row in both layouts.

…ts + generator)

…g (was resolving too high)

…reen (drop trailing goBack)

…; crypto flow self-cleans its output modal

…D (avoid last-tab interference), guard own-feed profile

…sion caps + tighter waitForExist polling

…lti-it flows; detect transitions, faster session caps, tighter polling

…l iOS e2e (test:e2e:ios)

Deletes .maestro/e2e flows/subflows/config, run-ios-devices.sh, generate-ios-report.mts
and the maestro functional yarn scripts. Keeps .maestro/performance and .maestro/visual-diff.
Renames the appium scripts to the canonical test:e2e:ios / :report. Updates the e2e skill.

…d per-device report

Parameterizes the appium port so each device runs its own server concurrently.
Caveat documented: macOS render-throttles backgrounded sim windows, so on a single
Mac parallel can flake/not beat serial — real speedup needs separate machines or
headless/cloud sims. Serial runner (run-ios-appium.sh) stays the reliable default.

…orientation, tablet-aware settings-subpages back nav

…tNav; bump mocha timeout

Tablet landscape LeftNav puts lower settings rows below the fold; scroll them
into view (isDisplayed-based) before tapping. Phone stays 16/16. Tablet Files
settings / files-folders remain flaky under the full suite (KBFS/two-pane timing).

…s); scroll member list for the smoke user

Keyboard from a prior input (feedback/chat/crypto) covered below-fold LeftNav rows
on iPad → unhittable. Dismiss it (guarded by isKeyboardShown — hideKeyboard hangs
with no keyboard). team-member scrolls the lazily-loaded member list to find the user.

…ne-session suite

…on 8100

Both sims defaulted to wdaLocalPort 8100; the second WDA collided with the
first (ECONNREFUSED 127.0.0.1:8100, sessions died mid-suite). Derive
wdaLocalPort/mjpegServerPort from the per-device appium port. Reuse the shared
prebuilt WDA (no derivedDataPath override — a fresh dir has no build to launch,
xcodebuild exits 65). Verified: iPhone+iPad both 16 passing in parallel.

…tandalone

yarn test:e2e:ios:report ran the generator with no env, falling back to the
single default dir (last single run = iphone only). Now scan tests/results for
ios-appium-debug-* and build one section per device, so the standalone report
covers iPhone+iPad. Falls back to the lone default dir when none exist.

testMatch globbed every *.test.ts, so yarn test:unit tried to run the appium
e2e specs (browser globals, no jest env). testPathIgnorePatterns only listed
the electron subdir; broaden to all of tests/e2e to cover ios-appium too.

- iPhone/iPad runs always write tests/results/ios-appium-debug-{iphone,ipad};
  report reads only those two dirs (env-var/auto-discovery logic removed)
- test:e2e:ios runs both devices in parallel; :iphone/:ipad run one serially
- per-image timestamp overlay (PNG mtime) and (stale) section marker (>1h)
  expose mixed-age runs; header links jump to each device section