fix(desktop): health check fails during server startup causing IDE instability

[herjarsa]

Pre-flight Checks

• I have searched existing issues and this is not a duplicate
• I understand this issue needs status:approved before a PR can be opened

Bug Description

The desktop app health check does not retry when the local server sidecar is still initializing. This causes:

1. MCP local connections to repeatedly disconnect and reconnect
2. IDE freezes when switching between sessions
3. Server marked as unhealthy even though it is still starting up

Steps to Reproduce

1. Open OpenCode desktop app
2. Server sidecar starts but health check fires immediately
3. Health check fails because server is not ready yet
4. Desktop app shows server as down
5. IDE becomes unresponsive or MCPs disconnect

Expected Behavior

Health check should retry with exponential backoff until the server is ready, up to a reasonable timeout.

Actual Behavior

Health check fails instantly on first attempt, causing cascading instability.

Operating System

Windows

Agent / Client

OpenCode Desktop

Shell

PowerShell

Relevant Logs

Server stdout shows "sidecar process spawned, waiting for server to become healthy" followed immediately by health check error without retries.

Additional Context

Related PR: #24138 which implements retry logic.

[github-actions]

This issue might be a duplicate of existing issues. Please check:

• [BUG] Windows: Sidecar handle leak causes desktop freeze + persistent server polling errors #23734: Windows sidecar handle leak causing desktop freeze + persistent server polling errors from startup (similar symptoms: Windows desktop instability, MCP disconnects, server errors during sidecar initialization)
• Desktop Electron can stay stuck on "Just a moment..." after sidecar is ready #23308: Desktop stuck on "Just a moment..." after sidecar is ready (related: health check / sidecar readiness detection failing to advance)

[herjarsa]

This is not a duplicate. The related issues (#23734, #23308) describe symptoms (desktop freeze, stuck on loading) that share root causes with this, but this issue specifically identifies the missing retry logic as the source.

PR #24138 fixes this by adding exponential backoff to the health check system.