fix(dashboard): retry stalled events feed reconnects

[helix4u]

What does this PR do?

Prevents the Dashboard events feed reconnect loop from stalling permanently after it has already displayed reconnecting in 1s....

The existing reconnect path only scheduled another attempt after a WebSocket close event. In gated Dashboard mode, every attempt first awaits buildWsUrl() to mint a fresh single-use ticket. If that request rejected or never settled, no replacement socket existed to emit close, so the retry loop stopped at its previous banner. A WebSocket handshake that emitted neither open nor close had the same failure mode.

This change gives the complete connection attempt, including ticket minting and the opening handshake, a 15-second deadline. Setup failures and deadline expiry feed back into the existing guarded exponential-backoff scheduler. A generation guard prevents late ticket responses or socket events from creating a superseded connection, and cleanup invalidates the active generation and clears both timers on unmount.

Related Issue

Discord support report: https://discord.com/channels/1053877538025386074/1535696700121948190

Follow-up to #79524.

Type of Change

• 🐛 Bug fix (non-breaking change that fixes an issue)
• ✨ New feature (non-breaking change that adds functionality)
• 🔒 Security fix
• 📝 Documentation update
• ✅ Tests (adding or improving test coverage)
• ♻️ Refactor (no behavior change)
• 🎯 New skill (bundled or hub)

Changes Made

• web/src/components/ChatSidebar.tsx
  • Catch URL/ticket construction failures and reschedule them through the existing reconnect policy.
  • Bound ticket minting plus the WebSocket opening handshake.
  • Invalidate timed-out attempts so late results cannot create competing sockets.
  • Clear the connection deadline on open, close, and unmount.
• web/src/lib/events-reconnect.ts
  • Define the events-feed connection-attempt deadline alongside the existing reconnect policy constants.
• web/src/components/ChatSidebar.test.tsx
  • Cover rejected URL construction, a stalled URL request with a late result, and a stalled WebSocket handshake.
  • Keep the existing error-plus-close single-scheduler regression explicit under the new handshake deadline.

How to Test

1. Run cd web && npx vitest run src/components/ChatSidebar.test.tsx.
2. Run cd web && npx tsc -p . --noEmit.
3. Run cd web && npx vitest run.
4. In a gated Dashboard session, interrupt the events-feed ticket request or opening handshake. Confirm the banner advances through backoff attempts instead of remaining at reconnecting in 1s... indefinitely.

Checklist

Code

• I've read the Contributing Guide
• My commit messages follow Conventional Commits (fix(scope):, feat(scope):, etc.)
• I searched for existing PRs to make sure this isn't a duplicate
• My PR contains only changes related to this fix/feature (no unrelated commits)
• I've run pytest tests/ -q and all tests pass (not run: this is a web-only TypeScript change; the full Python suite is intentionally left to CI)
• I've added tests for my changes (required for bug fixes, strongly encouraged for features)
• I've tested on my platform: Windows 11

Documentation & Housekeeping

• I've updated relevant documentation (README, docs/, docstrings) — N/A, no user-facing workflow changed
• I've updated cli-config.yaml.example if I added/changed config keys — N/A, no config keys changed
• I've updated CONTRIBUTING.md or AGENTS.md if I changed architecture or workflows — N/A, existing reconnect architecture retained
• I've considered cross-platform impact (Windows, macOS) per the compatibility guide — browser timer and WebSocket APIs only
• I've updated tool descriptions/schemas if I changed tool behavior — N/A, no agent tools changed

Screenshots / Logs

Validation from the clean worktree:

• Focused ChatSidebar.test.tsx: 16 tests passed.
• Full web Vitest suite: 27 files, 194 tests passed.
• Web TypeScript check: clean.
• ESLint on the three changed files: 0 errors and one pre-existing react-refresh/only-export-components warning at ChatSidebar.tsx:104.
• git diff --check: clean.

[austinpickett]

Superseded by #81978.

Your diagnosis was right and the fix is carried there as the first commit, unchanged — the deadline around ticket minting plus the handshake, the generation guard against a late ticket opening a superseded socket, and all three tests. EVENTS_CONNECT_TIMEOUT_MS living next to the other policy constants in events-reconnect.ts was the right call.

The reason for the salvage rather than a straight merge: the bug class is "the await buildWsUrl() in front of new WebSocket() is unguarded, so a rejected or hanging ticket leaves nothing to emit close" — and that structure exists on ChatPage's PTY socket too, which is the main chat surface. PTY_CONNECTING_TIMEOUT_MS doesn't cover it, because that timer is armed after new WebSocket(url) returns; a ticket that never settles never arms it, and connectInFlightRef stays stuck true, which also blocks the page-resume recovery path. #81978 applies the same shape there and adds the mirrored tests.

You're credited via Co-authored-by on the salvage commit. Thanks for chasing this one down to the actual mechanism — the write-up on the ticket-minting gap is what made the sibling call path obvious.