[ryanbr]

Summary

Found reviewing a 2026-09-28 strap log. Its alarm block showed Strap reports: 2045-06-10 14:03 and no Readback frame: line at all, and establishing why cost a trip through git history.

The frame line was emitted only when the frame existed. A readback carrying an epoch and no frame therefore printed nothing, which reads as "the frame was checked and was fine" rather than "there is nothing to check". Worse, the two causes are indistinguishable from outside the app:

• alarm: attribute the readback to a strap before judging it #1707 began banking the frame on 2026-08-28. Any readback taken before that has an epoch and no frame, permanently.
• A write that failed looks exactly the same.

The readback in that log was taken 2026-08-26, two days before the capture existed. That is knowable only from the repo, not from the export, which is the wrong way round for a diagnostic whose whole job is to separate a genuinely-stored stale alarm from a misdecode of a fixed response field.

What changes

The line now always appears. When there is no frame it names both possibilities instead of staying silent, the way the strap-clock verdict a few lines above it already says why it abstains rather than printing nothing:

Readback frame: not stored (this readback predates the frame capture, or the write failed), so a
genuinely-stored stale alarm cannot be told from a misdecode of a fixed response field here

Worth noting for whoever reads this next: the writer stores at least "empty" for any frame it decodes, so this branch is reachable in practice only for records banked before the capture existed. It will stop firing for a given install once its next WHOOP 4.0 readback lands. That is the intent, not a gap.

Parity

Both platforms carry it. The Apple twin also had a blank-handling divergence in the same expression: Kotlin used isNotBlank() while Swift used !raw.isEmpty, so a whitespace-only value printed a frame line carrying no bytes on Apple and counted as absent on Android. Swift trims first now, so blank means not stored on both.

Verification

• ./gradlew :app:testFullDebugUnitTest --no-build-cache --rerun-tasks: 763 suites, 6497 tests, 0 failures, result XML checked fresh against the clock.
• ./gradlew :app:lintVitalFullRelease -PstagingRelease: clean.
• 4 new Robolectric tests on Android and 4 Swift twins, both exercising the real alarmLines rather than asserting on source text. The Apple half is where the behaviour changed twice (the new branch, and blank handling), so it needed its own coverage; alarmLines there is a static func over UserDefaults.standard, seedable the way StrapProvidesLineTests already covers sibling lines.
• The Swift absent-frame assertion pins the Kotlin text verbatim, not a substring. The two strings are byte-identical today and these lines exist so an Android report and an Apple one compare directly, so asserting each side against only itself would let them drift apart silently.
• Mutation-checked on Android: removing the branch turns exactly the two new-behaviour tests red (absent frame, blank frame) and leaves both controls green (frame printed verbatim, no readback emits no line).
• python3 Tools/doc_comment_lint.py: clean.
• The Apple half is verified by the macOS and iOS legs rather than locally: this host is Linux.

Two things re-review caught, recorded because each would otherwise have shipped. The mutation run had the Kotlin branch deleted on disk when I re-read the diff, so the working tree looked like the fix was missing; judging a tree mid-mutation is as unsafe as editing one. And the Apple half had no tests at all, on the side where the behaviour changed twice.