Tests 1, 8, 9, 10 — measured here
This panel is the whole of test 10. Store 40 photos, press Mark day 0, come back in a week and read the verdict.
Storage
Environment
Add a note
Anything the numbers cannot capture — "permission prompt appeared again on launch", "torch had to be turned on from Control Centre".
Tests 2, 3, 4, 5 — QR decode
Pick the condition first, then New attempt. The clock starts on that tap, so camera warm-up is not counted against the decode. If it will not read, press Give up — a recorded failure is the point, not an embarrassment.
iOS needs a tap to open it.
Attempts this session
Tests 6, 7, 8 — batch capture and the photo pipeline
Take 10 without leaving this screen (test 6). The page watches the track and counts any teardown itself, so "the viewfinder never tore down" is a number rather than an impression. Photograph a bag of screws for test 7.
iOS needs a tap to open it.
Filling to 40 (test 8)
Real captures are better evidence than copies. If you run out of patience, padding is honest as long as it is recorded — the export labels padded rows separately.
Test 7 and 9 — do the photos survive, and are they good enough?
Thumbnails come out of IndexedDB, so this grid rendering with the network off is test 9. Tap one to open the 1600 px display tier and try to count the screws — that is test 7.
Export
Copy this into spike/RESULTS.md. Everything below was
measured on the device; nothing here is an impression.