ProjectSidewalk/SidewalkWebpage

When /routes or /stories has more than ListingMax (500) entries, the view renders <p class="community-cap-note"> saying the list is capped. frontend/css/pages/community-list.css colors it var(--color-neutral-600) (#8F8F8F) on the white page, which is 3.07:1 — below the 4.5:1 WCAG 2.1 AA minimum for small text, and the a11y: /routes meets WCAG 2.1 AA e2e spec fails on it: color-contrast [serious] .community-cap-note — Elements must meet minimum color contrast ratio thresholds CI never sees it because the CI database has far fewer than 500 routes; a dev database with a real city's routes (Seattle has ~1,950) shows it on every run. Darkening the note to a token that passes on white (--color-neutral-700 or the body text color) fixes it. Found while running the e2e suite for #5651. 🤖 Claude Fable 5.1 · effort: high

/signOut needs a session (SecuredAction), so a request without a cookie is sent to /anonSignUp first. That makes a new anonymous account, comes back to /signOut, and signs the account straight out again. Crawlers following the sign-out link do this on every visit. #5612 deleted 84,610 of these accounts on prod (2026-10-07), and the newest was from the day before, so they're still being made. Fix With no session, /signOut should just redirect to url. UserAwareAction instead of SecuredAction in UserController.signOut, logging SignOut and discarding the authenticator only when there is a user. 🤖 Claude Opus 5.5 · effort: medium

Every Gallery card renders the same ids, so a page of nine cards has nine copies of each: SeverityDisplay gives each filled circle id="current-severity" (public/js/gallery/src/displays/SeverityDisplay.js), styled by #current-severity in public/css/pages/gallery/cards.css. ValidationMenu builds #gallery-validation-button-holder and #gallery-card-agree-button / -disagree-button / -unsure-button per card, styled by their id rules in cards.css (both the base colors and the .gallery-card--readonly overrides). Duplicate ids are invalid HTML and make getElementById on any of them meaningless. The fix is classes for all of them, but note the specificity trap before flipping the CSS: the per-button id rules (1,0,0) currently beat .gallery-card:hover .validation-button-selected (0,3,0), which is why a selected card button keeps its own green/orange/grey background rather than --validation-button-selected-bg on hover. Class-based rules would lose that fight, so the per-button rules either need

The map status pill (public/js/ps-map/MapStatusPill.js, from #5002) shows "Zoom in to see the labels" on LabelMap and "Zoom in to see the clusters" on the AccessScore tool, then fades after 6 s. It has no X. The site's shared toast (public/js/common/Toast.js) sets the convention: auto-fade plus a close button, and hover or focus pauses the fade. The pill should match — a reader who has read it should be able to dismiss it, and one who is reading it slowly should not have it fade under them. Proposal Add a .map-status-pill__close button (the shared × styling, aria-label from common:close) to the pill; a click hides it at once. Pause the 6 s hint timer while the pill is hovered or holds focus, as Toast does. The "loading" state stays as is — it is a brief status, not a hint, and clears itself. Consider whether the pill should simply be a Toast (compact, dark variant) anchored to the map container, so there is one toast to maintain. Toast is fixed-positioned on <body>; the pill sits

Cleanup in test/service/ImageryFreshnessReportServiceSpec.scala:49 is a blanket delete by job name, run in both beforeAll and afterAll: private def cleanUp(): Unit = { val _ = run(sqlu"DELETE FROM background_job_run WHERE job_name IN ($pollJob, $syncJob)") } It removes every row under those two job names, not just the ones the spec seeded. On CI's empty schema that is the same thing. On a developer's database it is not: the dev app's own nightly actors record real runs under exactly those names, and running this spec destroys that history — which is what the Health panel and /adminapi/imageryFreshness read. It would also take a concurrently running suite's rows, and the beforeAll half means the loss happens even on a run that is interrupted before afterAll. Noticed while tracking down six orphaned check-imagery-age-actor rows in a dev DB (residue of two interrupted ImageryAdminSpec runs, unrelated to this). Those rows would have silently disappeared the next time this spec ran, whi