[Inbox scope] Keep metadata and disclosure honest across route transitions

[Chris0Jeky]

Context

Post-merge review of #2017 / #2018 confirmed two Inbox scope-transition races in the shipped visible-scope slice:

1. A late failure from obsolete board A metadata can clear successfully loaded board B metadata, leaving raw IDs.
2. Show all captures clears the scope disclosure before the unfiltered request commits, while the store deliberately retains the old board-only rows. Those rows can therefore appear to be the full Inbox; a failed unfiltered request can leave the misleading partial list visible.

Review threads:

• Disclose visible Review and Inbox scope #2018 (comment)
• Disclose visible Review and Inbox scope #2018 (comment)

Acceptance criteria

• Only the metadata request for the currently active board may update or clear scoped board/column names.
• Scoped names render only when their IDs match the current route, so an A-to-B route change cannot label B rows with A metadata while B loads.
• When clearing board scope, old board-scoped rows remain explicitly disclosed as scoped or are hidden/marked loading until the unfiltered list commits.
• A failed unfiltered request cannot present retained board-only rows as the complete Inbox.
• Add deterministic deferred-promise tests for the obsolete metadata failure and scoped-to-global list transition/failure orderings.
• Preserve the current route query/hash behavior and latest-request-wins capture-store contract.

Scope

Frontend-only Inbox orchestration and disclosure behavior. No capture API, persistence, or backend filtering change.

[Chris0Jeky]

PR #2043 is a bounded partial for this issue and remains non-closing.

At review-fix head d8544863bbb1b34cbe70f5299f1025790170e452 it:

• generation-guards board metadata and binds names to the active route IDs;
• masks retained rows/count in the table during route-scope replacement and failure;
• preserves an open child-local edit draft while that retained subtree is hidden (the P1 review defect is fixed and covered by DOM-identity/value/single-fetch regression evidence);
• passes 172 focused tests, typecheck, build evidence from the original head, and a fresh exact-head independent review with no HIGH/CRITICAL finding.

Two nonblocking residuals keep this issue open:

1. The page-level Inbox eyebrow counts still derive from retained prior-scope items during replacement; this is the exact remaining disclosure seam for this issue: Keep Inbox scope state honest across route transitions #2043 (comment)
2. Same-scope refresh failures should show still-valid rows beside the error; that distinct slice is tracked in Keep valid Inbox rows visible after same-scope refresh failures #2044.

Exact resume for this issue: mask or scope the page-level counts from the same replacement-state signal without hiding honest same-scope retained counts; add a deferred route-transition/failure test. Hosted CI and merge disposition for #2043 are still pending, so no acceptance box is being inferred complete yet.

[Chris0Jeky]

Cross-link from the alpha product-trust lane (2026-09-04): the 'page-level counts during replacement' residual that docs/STATUS.md parks on this issue is being implemented under #2501 (count-free Inbox eyebrow while isScopeReplacement is true, new inbox.eyebrow key in en/it/es) in worktree codex-1984; the PR will carry Refs #2022 and state exactly which of this issue's criteria it closes. No other part of #2022 is claimed.

[Chris0Jeky]

[Claude lane release v2]
lane: alpha-product-trust
issue: #2501 (and #2022, closed by the same PR)
PR: #2584, exact head 4648303, merge commit 44d50c9 (2026-09-05T02:34:52Z), merge commit not squash
disposition: closed by the merge. The Paper Inbox eyebrow renders a count-free label while a scope replacement is in flight (also on first mount, by design), captureStore.fetchItems reports whether it applied the response and the orchestrator clears the replacement flag only on an applied response, and a same-scope refresh with retained rows shows a visible refreshing note. #2022's two recorded residuals are the page-level counts (this PR) and same-scope refresh failures (#2044, shipped by #2484).
review: one fresh-context independent reviewer subagent; round 1 findings fixed in round 2; re-proved on the refreshed base after #2580 landed (combined Inbox set 9 files / 268 locally, then hosted CI at the exact head: CI, CI Extended, Smart CI green).
NOT verified: Playwright; the race end to end over a real network.
tracked non-blocking: #2591 (a dropped replacement response latches the flag with no retry affordance; reachable today only through Legacy's batch refresh).
next in lane: #2570 (was queued behind this PR) is now unblocked.
CLAUDE_SYNC_PACKET: merged=#258444d50c9 closes=#2501,#2022 refs=none docs=STATUS twelfth block (coordinator, pending) worktree=removed after this comment