Feature quality audit and mobile UX remediation across the whole platform

Why this audit

Some visible defects had survived multiple update cycles - most famously an old placeholder loading screen on Android. Instead of patching symptoms again, this pass audited every surface (four mobile apps, the web consoles, the public site, the backend, payments, dispatch and tracking, store readiness, tests and operations) and asked one question first: why do these keep coming back?

The loading-screen story, root-caused

• The 'target/bullseye' start-up screen was literally template placeholder art that had been shipped as the splash logo in all four Android apps, at every screen density.
• It kept surviving because the native Android asset directory lives outside version control and code review, and because a build helper could silently reuse stale build artifacts.
• Fix: brand marks are now generated per app from committed sources, a guard in the standard checks refuses the placeholder art forever, the artifact helper refuses stale builds, and a white flash between the splash and the first app frame was eliminated.
• Proof: fresh builds installed on a physical Android 15 phone show each app's own branded splash and a seamless dark launch - screenshots are part of the internal evidence pack.

Distinct app icons

Customer and Courier shared one launcher icon, Merchant and Fleet shared another - the same class of issue that app stores reject. Each app now ships its own mark (bag, cap, storefront, truck), regenerated at every density and protected by an automated duplicate-icon check.

Payment pipeline correctness

• Fixed a defect that prevented signed-in customers' card payments from completing, and added regression tests that exercise the exact production configuration.
• Hardened the payment-webhook recovery path so a card charge that fails and then succeeds on retry always lands correctly, even under the strictest order-state enforcement.
• Late or duplicate failure events can no longer touch an order that is already being prepared or delivered.
• The health endpoint now reports exactly which build is running in production, so every future deploy is verifiable.

Production data guardrails

Demo and test tooling is now hard-walled from production data: destructive test-data operations refuse to run in the production environment unless an explicit, temporary operator override is set. Internal alert routing was also corrected so partner applications always reach a real inbox.

Verification

• Real device: all four apps installed and cold-start-verified on a physical Android 15 phone (splash, icons, launch, core screens).
• Web: 9 public pages x 4 viewport widths - zero console errors, zero failed requests, no layout overflow.
• Tests: the full suite (195 files, 1,440 tests) is green, including 9 new regression tests written for the defects fixed here.
• Method: 18 specialized AI reviewers, an independent synthesis, and an adversarial council that re-verified every priority claim in code before anything was changed.

What comes next

• Publishing the corrected app binaries to the store track so every installed phone receives the new launch experience.
• Push notifications, crash reporting and session persistence for the mobile apps - audited, scoped and queued.
• Courier navigation handoff, structured opening hours, and richer merchant order detail - part of the prioritized remediation backlog produced by this audit.