SSH readiness proof — real Linux Electron desktop
AI-assisted; verified with the tests and tools listed above.

Fixture: dedicated OpenSSH server, synthetic account, published remote t3 0.1.9.
Network: 500 ms netem egress delay at each veth endpoint.
Measured ping RTT: min/avg/max 1000.056/1000.069/1000.077 ms, zero loss.

Control during the base failure (same remote server):
  Remote loopback GET /: HTTP 200, 0.001629 s.
  GET / over a separate real ssh -L tunnel: HTTP 200, 3.003298 s.

Base desktop reports:
  httpReadiness.timedOut: timeoutMs=20000, probeTimeoutMs=1000.
  ssh.tunnel.ready.failed: processRunning=true, localPortListening=true.
  Cause: Backend readiness probe exceeded 1000ms.
  UI: SshReadinessError: Timed out waiting 20000ms for backend readiness.

Head desktop reports:
  ssh.tunnel.ready, followed by ssh.environment.ensure.succeeded.
  remoteServerKind=managed; issuedPairingToken=true.
  UI eventually settles at Connected.

Observed remaining delay on head:
  The follow-up existing-tunnel check has timeoutMs=2000 even with
  probeTimeoutMs=5000. It timed out after one attempt, treated the tunnel as
  stale, and recreated it before the client reached Connected.
  This proof demonstrates eventual opening, not seamless tunnel reuse.

Final recordings, from Add environment click to observed outcome:
  Base: readiness error at 58.156 s; stable identical stills 500 ms apart.
  Head: saved environment Connected at 95.906 s; stable identical stills 500 ms apart.
  Both videos: 1100x780, 10 fps, recorded in real time with no acceleration.