MOR-2436: compact the Standard VFO pair bridge

[morozsm]

The installed Standard VFO pair grew to 246px when the complete IC-7300 shared facts and operations were present, leaving a large empty strip below both receiver cards. This compacts the center bridge into bounded grids and lets the active status strip wrap inside the card's reserved second line. The production-shaped browser regression measures a 185px row with 183px aligned cards at 1200px and 1700px, with every selector, SPLIT, A=B, A↔B, SPEAK, RX/TX digest, and card chip visible without overflow. At 900px the pair reflows while preserving every chip and leaves a measured 1.75px gap before the station meters.

This PR also carries the reviewed MOR-2426 reconnect correction. The original 51-query connection refresh could age the PTT observation until there was no admissible confirmed-RX proof for the select-A policy. When RX proof is stale or unavailable, the connection policy performs one bounded passive PTT refresh before selecting VFO A; fresh confirmed RX skips that read, and fresh TX skips the selection policy. Selection still fails closed unless the available observation reaches canonical ingress as confirmed RX.

Linear: MOR-2436, MOR-2426

Validation:

• npm run check
• npm run lint
• npm run build
• npm run test:e2e:i18n -- tests/e2e/i18n/desktop-geometry.spec.ts (46/46)
• focused 900/1024/1200/1700 compact-pair and 900 observed/unknown geometry cases (6/6)
• Backend slice: 269 focused tests plus mypy, ruff, and import checks (reviewed exact source commit 1db7a8c4bc0dc352087bff8602bffe099c9d4033)

[morozsm]

Agent Review: BLOCKED 6fb3d7b

The backend slice independently passes review (269 focused tests; source patch identical to reviewed 1db7a8c4bc0dc352087bff8602bffe099c9d4033), and the compact VFO pair passes independent 1200/1700 geometry review with all 9px chips and central operations visible (185px row, 183px cards).

The full quick run 34432586881 exposed a remaining narrow-viewport acceptance failure: both standard 900 observed IC geometry and standard 900 unknown geometry fail desktop-geometry.spec.ts:592 because the scope bottom is 773.0625 while the station-meter boundary allows at most 761.8125. Diagnose and fix the overlap without weakening the assertion; rerun both affected cases and the complete i18n smoke suite. The remaining 89 cases passed, and visual CI passed.

Also correct the PR body's obsolete 167px measurement to the final measured height, and describe the installed reconnect failure as lacking admissible RX proof at that instant. PTT aging is a reproduced regression condition, not a captured raw live observation.

No deployment or merge until the updated exact head has independent delta review and required green CI.

[morozsm]

Agent Review: PASS 73b50e7

Independent combined code review passes. This supersedes the 6fb BLOCKED verdict: the narrow Standard overlap is corrected without weakening the scope/meters boundary assertion. Independent UI delta verification passes all six targeted cases, including full-rich absolute cards at 900/1024/1200/1700 and the two previously failing 900 cases. The 900 scope bottom is 759.0625 and meters begin at 760.8125, a positive 1.75px gap; descendant containment and screenshots show no clipped controls. Wide rows remain 185px with aligned 183px cards and retained 9px status chips. Builder's complete desktop geometry run passed 46/46.

I independently reviewed the backend admission/read/ingress/epoch paths and ran 269 focused tests successfully on its source slice. The backend patch in this final combined head is byte-identical to reviewed 1db7a8c4bc0dc352087bff8602bffe099c9d4033 (SHA-256 6746d81fd98498088d27be1e8b8975e6eb0bfcb10839148f7d473f40ba9d7912). Fresh TX/read-only/inapplicable paths skip selection; stale/unavailable RX proof permits only one bounded passive profile-declared read, and canonical fresh RX plus unchanged epoch are still required for the existing typed A select. Timeout, missing ingress, TX reply and generation churn remain closed.

PR prose now records the corrected geometry and regression evidence. At this review, visual CI is SUCCESS and canonical quick is running; required exact-head CI remains a separate pre-merge gate. Installed IC-7300 reconnect and final visual acceptance will follow deployment and are not claimed by this code verdict.