Telemetry audit and observability instrumentation

[EhabY]

Problem

We have zero external observability. The extension logs locally (VS Code output channel, HTTP/WebSocket loggers) but nothing leaves the client. We can't detect regressions after a release or measure connection reliability.

Instrument

Startup

• Activation success/failure (categorized)
• Time-to-activate
• Auth state at startup (valid token, expired, none)

Connections

• Time to first successful connect
• Connection drops: frequency, duration, cause
• Reconnection attempts: count, success rate
• Certificate refresh failures

Errors

• Error rate trends per release (version-tagged)
• Categorized: auth, network, certificate, build, agent
• Unhandled promise rejections

Investigate

• Sentry + source maps for client-side crash reporting and stack traces
• VS Code TelemetryLogger API — respects user telemetry preferences out of the box, may be sufficient
• Privacy review — ensure no deployment URLs, tokens, or workspace names leak into telemetry

Approach

1. Add a telemetry abstraction so we can swap backends later
2. Respect VS Code telemetry preferences (vscode.env.telemetryLevel) from day one
3. Instrument highest-value signals first: activation failures, connection drops, error rates