Fix Orchestrator task persistence and file lock handling

[roomote]

Fixes the issue where Orchestrator mode tasks would stop prematurely and disappear from the task list after creating multiple subtasks in rapid succession.

Root Cause

Lock contention during rapid file writes when Orchestrator creates multiple subtasks. Each delegation involves concurrent writes to:

• Parent task API conversation history
• Child task API conversation history
• Task history metadata

When lock acquisition failed or metadata persistence failed silently, parent tasks would appear to "disappear" from the UI.

Changes

1. Increased lock resilience in safeWriteJson:

   • Lock staleness timeout: 31s → 60s (handles complex orchestrator scenarios)
   • Lock retries: 5 → 10 attempts
   • Backoff timeout range: 100-1000ms → 200-2000ms
   • Enhanced error messages with contention context

2. Added retry logic for critical delegation metadata:

   • 3 retry attempts with exponential backoff for parent task status persistence
   • User-visible warnings when all retries fail (prevents silent data loss)
   • Detailed logging for debugging lock contention issues

Testing

• All existing safeWriteJson tests pass (16/16)
• All delegation tests pass (13/13)
• No lint errors or type check failures

---

View task on Roo Code Cloud

---

Important

Enhances task persistence and lock handling in Orchestrator mode by adding retry logic and increasing lock resilience in ClineProvider.ts and safeWriteJson.ts.

• Behavior:
  • Fixes premature stopping and disappearance of Orchestrator tasks after creating multiple subtasks.
  • Adds retry logic for parent task metadata persistence in ClineProvider.ts.
  • Increases lock resilience in safeWriteJson.ts with longer timeouts and more retries.
• Lock Handling:
  • safeWriteJson.ts: Lock staleness timeout increased from 31s to 60s, retries from 5 to 10, and backoff timeout range from 100-1000ms to 200-2000ms.
  • Enhanced error messages for lock acquisition failures.
• Retry Logic:
  • ClineProvider.ts: 3 retry attempts with exponential backoff for parent task status persistence.
  • Logs user-visible warnings and detailed debugging information when retries fail.
• Testing:
  • All existing safeWriteJson tests pass (16/16).
  • All delegation tests pass (13/13).

^{This description was created by for 8a5eccb. You can customize this summary. It will automatically update as commits are pushed.}

[roomote]

Rooviewer   See task on Roo Cloud

Review complete. No issues found.

The changes appropriately address lock contention during rapid orchestrator delegation by:

• Increasing lock resilience parameters in safeWriteJson (staleness timeout, retries, backoff ranges)
• Adding retry logic with exponential backoff for critical parent task metadata persistence
• Providing user-visible warnings when all retries fail instead of silent failures

_{Mention @roomote in a comment to request specific changes to this pull request or fix all unresolved issues.}