Answer Apple's 2.1 request, fix three bugs found on the bike, and measure what riding apps really send

[sbroenne]

Apple rejected 1.0 under Guideline 2.1 and asked for a screen recording made on a
physical device, plus written answers. The written answers are already live in App
Store Connect; this branch carries the shot list for the recording — and everything
found while getting a phone ready to shoot it.

Ready for Apple

• docs/DEMO_VIDEO.md is a shot list that can actually be filmed: launch and
  connect, shifting on the real trainer, then Demo Mode. No pedalling, no riding
  app required. It also holds the reply text, corrected to stop claiming the video
  shows pedalling.

The video has not been recorded yet. Do not resubmit before it exists.

Three bugs, all found on the bike and all confirmed fixed there

• The Zwift Click was never found. Discovery ran once for three seconds and then
  stopped listening for good. A Click sleeps and only advertises briefly after a
  button press, so that window had almost always closed first. It now keeps looking.
• Every device state vanished while the riding app was disconnected. The footer
  swapped itself for a single problem line whenever anything was not .ok, and
  "waiting for the riding app" — normal for the first minute of every session — was
  enough to hide the trainer and the Click. Only real faults do that now.
• Waking the Click shifted a gear. Pairing one means pressing a button, and it
  often connects with that button still down. The edge tracker read the first report
  as a fresh press: a gear moved on its own, and half a second later the hold began
  and kept moving them. The first report after connecting is now only a baseline.

What the hardware actually said

Three measurements, with their logs, in DEVELOPMENT.md:

• RealVelo never sets a wheel size, across a whole ride and 147 terrain updates.
  FulGaz sets one at every ride start. The app cannot assume either.
• A riding app's reset does not wipe the wheel size. RealVelo sends one while
  connecting, which by the FTMS rules should return the trainer to its defaults and
  silently undo the current gear. Measured on a real KICKR V5 rather than argued
  about: leave a size nothing defaults to, send the same reset, work out what the
  trainer is riding on. It survives, so passing resets through is safe.
• The trainer keeps an odd wheel size across a power cut. Recorded for the first
  time. Putting the wheel size back after a crash genuinely matters; nothing else
  will do it.

The probe could not wait

KickrProbe.run() executes inside settleTask, and subscribing to the speed channel
part-way through fired the callback that cancels it. A cancelled task's sleeps return
at once, so every timed wait collapsed and the pedals were never given a chance to
turn. Fixed, and waiting is patient now — the person doing this has to walk to the
bike, and a deadline only makes them race the tool. Hardware evidence recorded before
this fix should be treated with suspicion; the power-cut figure above was re-measured
after it.

AppTap now calls out anything unusual as it happens rather than leaving it buried in
a list of opcodes, and run-tool.sh kills a previous instance of the same tool, after
three AppTaps once appeared in one pairing list at the same time.

Still outstanding

• The video is not recorded and nothing has been resubmitted to Apple.
• Nothing on this branch has been ridden. The fixes above were confirmed on the bike
  without pedalling.
• Zwift and MyWhoosh have not been traced.

Resolves #1

[Copilot]

Pull request overview

Updates the App Store submission documentation to align with Apple’s Guideline 2.1 “Information Needed” rejection, specifically clarifying that a physical-device screen recording is required and providing an updated, shootable recording plan plus matching App Review Notes text.

Changes:

• Reworked docs/DEMO_VIDEO.md to make the recording mandatory, spell out Apple’s requirements, and rebuild the shot list around a realistic filming flow.
• Updated docs/APP_STORE.md Stage 6 Notes template to mirror the live App Review Notes and added a Guideline 2.1 row to the rejection-reasons table.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

File	Description
docs/DEMO_VIDEO.md	Makes the screen recording requirement explicit; defines Apple-required constraints; provides an updated shot list and App Review reply template.
docs/APP_STORE.md	Updates the App Review Notes copy and rejection table to reflect the 2.1 request and the now-required recording.

Suppressed comments (1)

docs/DEMO_VIDEO.md:146

• The reply template hard-codes the recording device and iOS version ("iPhone 17 Pro running iOS 26.6"), which will be wrong if the actual recording is captured on a different model or after an iOS update. Since Apple’s requirement is "latest OS", the reply should describe that generically and avoid forcing a specific device/version.

> Thank you. A screen recording made on a physical iPhone 17 Pro running iOS 26.6
> is attached. It starts by launching the app from the Home screen, shows the
> Bluetooth permission prompt, the app finding and connecting to a Wahoo KICKR V5

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.