Clarify Plan v8+ worker crash completion behavior in response to review feedback

[Copilot]

Context

Addresses review comment on #5412 regarding the intentional design decision to call ReportJobCompletionEventAsync instead of CompleteJobRequestAsync for Plan v8+ worker crashes.

Review thread: #5412 (comment)

---

Description

Provided technical explanation clarifying the completion flow for Plan v8+ worker crashes:

• Plan v8+ worker crashes: Listener calls ReportJobCompletionEventAsync to send plan event directly, since crashed worker cannot report completion
• Plan v8+ normal completions: Still calls CompleteJobRequestAsync, which returns early (lines 959-963) for Plan v8+ as plan events replace legacy FinishAgentRequest API
• Plan v7 scenarios: Always calls CompleteJobRequestAsync, which executes FinishAgentRequest

The conditional logic ensures job completion is properly reported regardless of worker state, leveraging Plan v8+'s event-based completion mechanism when available.

if (isPlanV8Plus && isWorkerCrash)
{
    // Worker crashed - listener must send plan event
    await ReportJobCompletionEventAsync(message, result);
}
else
{
    // Normal flow - CompleteJobRequestAsync handles both v7 and v8+
    await CompleteJobRequestAsync(_poolId, message, lockToken, result, detailInfo);
}

---

Risk Assessment (Low)

No code changes - documentation/clarification only.

---

Unit Tests Added or Updated (No)

No code changes made.

---

Additional Testing Performed

Not applicable - response to review comment only.

---

Change Behind Feature Flag (Yes)

Existing behavior already behind AgentKnobs.EnhancedWorkerCrashHandling feature flag.

---

Tech Design / Approach

• Design previously reviewed in Enhanced worker crash handling with integrated crash telemetry #5412
• Clarification confirms intentional behavior, not a bug

---

Documentation Changes Required (No)

Code comments already document the conditional completion strategy.

---

Logging Added/Updated (No)

Existing trace statements cover all code paths.

---

Telemetry Added/Updated (No)

Existing telemetry unchanged.

---

Rollback Scenario and Process (Yes)

Feature flag allows instant rollback if needed.

---

Dependency Impact Assessed and Regression Tested (Yes)

No behavioral changes - review feedback addressed only.

---

💡 You can make Copilot smarter by setting up custom instructions, customizing its development environment and configuring Model Context Protocol (MCP) servers. Learn more Copilot coding agent tips in the docs.