Problem

The proxy startup logic in start-proxy-action.ts can report successful initialization even when the proxy process fails to stabilize after retry attempts.

The previous implementation:

• Retries on spawn or early exit
• Only throws on immediate spawn errors
• Does not explicitly fail when all retry attempts are exhausted
• May proceed with outputs set despite no stable process remaining alive

This can lead to false-positive startup.

Impact

When startup silently fails:

• Downstream steps may fail with opaque dependency or networking errors
• Failures manifest later in the workflow
• Root cause becomes difficult to diagnose

Instead of failing fast during initialization, workflows continue in a degraded state.

Expected Behavior

If the proxy fails to stabilize after retry attempts, startup should fail explicitly with a clear diagnostic error.

Proposed Solution

• Detect retry exhaustion explicitly
• Throw a clear error including the last exit code
• Verify process remains alive after stabilization delay
• Retry using a new ephemeral port before terminal failure
• Extract startup logic into a dedicated launcher module for testability

This preserves existing outputs, telemetry behavior, and wrapper semantics while preventing silent failure.