Telemetry: instrument CLI resolution and remote setup phases

[EhabY]

Follow-up from the local telemetry audit of #953 / #903.

Problem

Connection/setup support cases often hinge on CLI resolution, binary downloads, config writing, compatibility checks, websocket setup, and SSH process behavior. Current telemetry has useful top-level coverage (remote.setup, cli.download, SSH process lifecycle), but several slow/failure-prone phases are not separately visible.

Proposed scope

CLI binary resolution

Add a cli.resolve trace with child phases such as:

• cli.resolve.cache_lookup
• cli.resolve.version_check
• cli.resolve.lock_wait
• cli.resolve.download
• cli.resolve.verify
• cli.resolve.fallback

Useful properties:

• result: success, failure, cancelled
• resolved source: cached, downloaded, bundled, system, fallback
• requested/actual version, if safe and useful
• downloads disabled flag
• verification outcome
• bounded failure category
• duration

CLI configuration

Add telemetry for configuring the CLI for a deployment/workspace without recording secrets:

• cli.configure.started
• cli.configure.succeeded
• cli.configure.failed

Useful properties:

• config mode/category
• credential source category, not value
• result/failure category

Remote setup phases

Extend remote.setup child phases to cover more of the user-visible setup path:

• remote.setup.cli_resolve
• remote.setup.auth_setup
• remote.setup.compatibility_check
• remote.setup.workspace_monitor
• remote.setup.websocket_monitor
• remote.setup.ssh_monitor
• remote.setup.vscode_handoff

This should make it possible to answer: “where did connection setup spend its time?”

Websocket/reconnect diagnostics

Add bounded reconnect attempt telemetry:

• attempt number bucket
• backoff duration
• route bucket with query stripped
• result
• failure category

Avoid raw query strings or unnormalized paths.

SSH setup/process diagnostics

Add events/traces for setup points not currently visible:

• ssh.config.generated
• ssh.config.failed
• ssh.process.spawned
• ssh.process.spawn_failed
• ssh.handshake.failed
• ssh.port_forward.started
• ssh.port_forward.failed
• ssh.connection.stable

Useful properties:

• result/failure category
• duration
• process source/category
• no private key paths or raw command lines unless explicitly redacted

Acceptance criteria

• CLI cache hit, download, verify, fallback, and failure paths are distinguishable in telemetry.
• Remote setup traces show the major phases between command invocation and VS Code remote handoff.
• Websocket reconnect telemetry uses bounded, privacy-safe properties.
• SSH setup/process failure modes are bucketed and visible.
• Tests cover representative success/failure paths.

---

Generated by Coder Agent from the telemetry audit of #953.