[siriushex]

Changes

• Preserve original optional UTC offsets without shifting recorded UTC instants.
• Validate timestamp aliases and the emulated extended-bolus child's own date/offset. Contradictions invalidate the retained insulin inventory before time-window filtering, including through native storage, cache refresh and reopen.
• Mark Combo-labelled bolus timestamps as unverified delivery starts. Keep original doses and reasons; expose only active uncertainty/missing-offset counts through the existing owner.
• Document source semantics and remaining delivery-completeness and clinical-admission requirements.

Verification

• Failing-first pure and native Room regressions reproduced.
• Final frozen full Android test/lint/compile/assemble: 6815 passed, six existing private-input skips; all ten new cases executed.
• Real pinned stock AAPS SDK TLS test passed (2.708 s); zero lint errors; 1431 source hashes unchanged throughout the final run.
• APK retention check: all 7178 application classes retained; local build 40,670,995 bytes.
• Existing mixed-child regression fixture strengthened with a valid date and positive control so time validation cannot mask its intended assertion.

Boundaries

No AAPS, phone, server, target, dose or alert changes. No private data, logs, credentials or APK in this PR. Recommended meal time remains disabled: timestamps and successful synthetic tests are not proof of complete physical delivery or a clinically validated waiting interval. Source CI and final main CI are separate integration gates.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-23T11:06:34.682051Z	70f6be7	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.