[Bug]: Desktop: relay environment discovery never runs/retries when the app starts without network, remote T3 Connect environment stays "offline" until the next restart#17106

Before submitting

• I searched existing issues and did not find a duplicate.
• I included enough detail to reproduce or investigate the problem.

Area

apps/web

Steps to reproduce

Summary

My laptop came out of Modern Standby on a phone hotspot (5G), and the desktop app lost its connection to my remote T3 Connect environment. I then restarted the app while the network was still coming up. After that restart, the app never ran relay environment discovery. The remote environment stayed "offline" for 46 minutes, even though the server was healthy and reachable the whole time. Clicking Reconnect did nothing on the network.

Restarting the app again once the network was stable fixed it immediately. This time it was on a regular Wi-Fi network. Relay discovery ran 3 seconds after startup, cleared the stored offline report, and connected.

This looks like a client-side counterpart of #12061, where something that fails once at startup is never retried.

Server side was healthy

Checked at the same time:

• The service had been up for 3 days with 0 restarts, and the local origin /.well-known/t3/environment returned 200.
• cloudflared was running, with only routine re-registrations in the log.
• From the laptop on the same hotspot, relay.t3.codes returned 200, and the environment endpoint via a separate path (Tailscale) also returned 200.

Suspected cause

Likely in packages/client-runtime (relay/discovery.ts, state/relayDiscovery.ts), surfaced via apps/web/src/state/relay.ts in the desktop renderer.

At startup, the relay environment discovery (managedRelaySession → RelayEnvironmentDiscovery.refreshStatus) either fails before it is traced, or is skipped while the network/relay isn't reachable yet. Either way it is never scheduled again. The stored offline report for the remote environment is only cleared by a successful refreshStatus, so the environment keeps showing offline. retryNow for that environment doesn't trigger discovery, so Reconnect has no effect.

The earlier drop during standby (07:42 – 08:28) may be a separate issue: occasional relay requests succeeded but the websocket was never re-established. The trace for that period doesn't show why.

Steps to reproduce (best guess)

1. Connect the desktop app to a remote environment via T3 Connect.
2. Disconnect the network (or sleep the laptop and wake it on a network that takes a while to come up).
3. Start or restart the desktop app while there is no connectivity yet.
4. Restore connectivity. The remote environment stays offline, and Reconnect does nothing.
5. Restart the app with working network. The environment connects within seconds.

Related

• #12061: the same "fails once at startup, never retried" pattern for the host-side relay link (clock skew)
• #15051, #13266: the remote environment never connects because it is switched off. Ruled out here, the toggle was On
• #4589: subscriptions die after a transport blip until the client restarts

Expected behavior

• Relay discovery and the connection to the remote environment are retried with backoff, or when connectivity returns. A startup without network shouldn't leave the environment offline until the next restart.
• Reconnect actually re-runs discovery and starts a connection attempt.

Actual behavior

• After the 08:30 restart, the only spans in the trace for that whole period are:
  • EnvironmentSupervisor.make + ConnectionDriver.connect for PrimaryConnectionTarget;
  • the laptop's own managed tunnel recovery, which failed 6× and then succeeded at 08:32.
• No clientRuntime.managedRelaySession.readClerkToken, RelayEnvironmentDiscovery.refreshStatus or web.managedRelayDpopSigner.createProof between 08:30 and 09:16.
• Every Reconnect click logs EnvironmentRegistry.retryNow → EnvironmentSupervisor.retryNow → setState, and no request follows.
• The app keeps throwing "<env> is not connected." until 08:49.

Impact

Major degradation or frequent failure

Version or commit

Desktop app 0.0.45 (Windows); remote service t3@0.0.45

Environment

Windows 11 Home 10.0.26200, T3 Code desktop 0.0.45. Remote: Ubuntu Server, t3@0.0.45 as systemd user service via T3 Connect (managed cloudflared, TUNNEL_TRANSPORT_PROTOCOL=http2). Client network: phone hotspot (5G) during the failure, office Wi-Fi when it recovered. Remote environment toggle was On.

Logs or stack traces

Screenshots, recordings, or supporting files

No response

Workaround

No response