[ael-dev3]

Tall replies previews could overlap the phone’s two-row sticky header because placement reserved a fixed 72px. Measure the actual header bottom when opening anchored and nested panels. A controlled phone fixture reproduces the old overlap; the desktop case preserves its clearance checks.

The scheduled-cast edit fixture also selected the second occurrence of Berlin’s repeated 02:00 during tonight’s daylight-saving transition. Both UTC instants serialize to the same datetime-local value, which the browser parses as the earlier occurrence. Choose tomorrow’s local noon while retaining exact signed timestamp, signature, message, parent, replacement and status assertions.

Validation: required Rust checks and 607 native tests; fresh build; 16 focused reply-preview, nested-card, post-popover and scheduling cases; strict phone/desktop smoothness evidence recorded before publication. No thresholds, workflow triggers, provider configuration or skip rules changed. The preceding main CI and full 919-case production run remain recorded as failed; neither was rerun.

The first PR run was rejected because the Escape/re-entry fixture missed its short Closing state, then passed on retry. The follow-up replaces a whole-page clock installed after opening with a close-callback hold installed before the app loads. Real pointer re-entry and its dwell still run; the test requires Open before Escape, Closing before and after re-entry, one released close, no remaining panel, and a normal fresh hover. A controlled local restoration of the old reopening behavior fails this check; the unchanged product build passes nine adjacent cases and ten no-retry repetitions. The exact event sequence behind the original CI miss was not captured, and a late-clock wall-delay diagnostic passed locally; that failed CI remains retained rather than relabeled.