Auth/API timeout should keep users in-app with clear unsaved-work warning

[wilhel1812]

Summary

When the startup auth/profile request /api/me times out or returns a Cloudflare/server timeout, LinkSim currently renders a full-page signed-out/access-unavailable state. That can look like the user was thrown out of the app even when their Cloudflare Access session is valid.

Desired behavior

• Keep the user in the workspace for auth/API timeout, signed-out, forbidden, pending, and revoked states.
• Show a persistent, impossible-to-miss in-app warning that cloud save may not work and changes may not be saved.
• Reuse the existing AppShell notification stack/UI-gallery pattern; do not introduce a new banner system.
• Keep current edit/save permissions by default; this issue improves warning and continuity, not local draft persistence.
• Make /api/me fail fast with clear JSON errors instead of hanging until Cloudflare returns 524.

Acceptance criteria

• Startup /api/me timeout or 524 keeps the normal workspace visible.
• Auth degraded states show a pinned accessible warning.
• Full-page access lockout is removed for these auth states.
• Backend auth verification is bounded so /api/me does not hang indefinitely.
• Tests cover frontend degraded auth handling and backend auth timeout/header paths.

[wilhel1812]

Implementation starting after staging drift reconciliation. Baseline: origin/staging 7522b54, staging deploy completed with build info 0.18.0+7522b546.

[wilhel1812]

Implementation PR opened: #779. Local verification passed: npm test (92 files / 439 tests) and npm run build.

[wilhel1812]

Merged to staging via PR #779. CI staging deploy completed for commit 617a244 with build info 0.18.0+617a2443. Local and PR verification passed: npm test and npm run build.

[wilhel1812]

Starting second implementation batch for automatic auth recovery retries. Plan: quick retries after auth/bootstrap failure, then 60s background retries while degraded and online; no new visible UI controls.

[wilhel1812]

Implementation update: added the auth recovery retry iteration on branch issue/776-auth-recovery-retry. The AppShell access check now retries automatically after transient /api/me failures with 2s/5s/10s quick retries, then 60s background polling while online, and retries immediately on the browser online event. Intentional local read-only and unauthenticated deep-link guest flows do not enter recovery retry. Local verification passed: git diff --check, npm test, and npm run build.