FleetCockpitVisual Accounts test broken since the #15398 config-card re-skin

Context

test/playwright/visual/FleetCockpitVisual.spec.mjs → "the Accounts surface" test has been red on dev since #15398 / 0889cc99b1 (2026-07-18 04:46, the define-agent config-card re-skin) introduced fm-agent-config-card as the card's class: the spec's .agent-config-card selector (written for the older card in #15007, 2026-07-11) matched nothing from that merge on. #15503 / 91a0620d98 (the Accounts trio, same day 21:08) later changed the surface the refreshed golden represents (selector strip, harness chips, untruncated labels). The current syncAgentSelector() already selects the first fixture record, so no selection drive is needed — the class-name repair alone unblocks the wait.

The golden predates the restructure (captured 2026-07-11 in #15007), so this is spec rot, not a design regression.

Live latest-open sweep: checked latest 20 open issues at 2026-07-21 ~19:35Z; no equivalent found. A2A in-flight claim sweep (last 60 min): no competing claim.

The Fix

Repair the stale selector in the Accounts test (.agent-config-card → .fm-agent-config-card), then refresh accounts-config-surface.png under the visual config's refresh discipline (--update-snapshots, diff reviewed in the PR). Verified live: syncAgentSelector() already selects the first fixture record at boot, so no selection drive is needed — the class-name repair alone unblocks the wait.

Acceptance Criteria

• The Accounts test boots deterministically to the config card and goes green on current dev
• The refreshed golden reflects the current Accounts surface (selector strip + config card)
• No change to the other three FM cockpit baselines beyond their reviewed state

Related

• #15007 (baseline harness), #15489 / #15503 (Accounts trio — the restructure), #15649 (the lane that surfaced the rot)

Origin Session ID: d8a51237-4fcc-4171-8071-a391da0be361

Retrieval Hint: "Accounts visual test agent-config-card stale golden selector strip"