openclaw - 💡(How to fix) Fix Session file lock timeout when multiple subagent completion events arrive concurrently

Problem

When multiple subagents complete near-simultaneously, their completion announce events compete for the parent session's file lock, causing timeout and delivery failure.

Reproduction

1. Spawn 3 subagents in parallel from the same parent session (e.g. stuart-trader spawns 3 stuart-coder instances)
2. All 3 complete within seconds of each other
3. Gateway attempts to deliver completion events to the parent session concurrently

Observed Error

[warn] Subagent announce give up (retry-limit) run=96add4e4 child=agent:stuart-coder:subagent:dac1add9 requester=agent:stuart-trader:discord:channel:1505029661288042538 retries=3 endedAgo=124s deliveryError="completion agent handoff is still pending; direct-primary: completion agent handoff is still pending"

[agents/agent-command] Turn transcript persistence failed for agent:stuart-trader:discord:channel:1505029661288042538: session file locked (timeout 60000ms): pid=6079 ...sessions/e6ab114a.jsonl.lock

⚠️ ✉️ Message failed

Impact

• Completion events silently dropped after 3 retries
• Parent agent never receives some subagent results
• Downstream Discord message delivery also fails

Environment

• OpenClaw v2026.5.27
• 3 parallel subagents (openai-codex/gpt-5.5) spawned from one parent (github-copilot/claude-opus-4.6)
• Host: Ubuntu 26.04, Node v24.15.0

Suggestion

Consider a per-session write queue (FIFO) or serialized lock acquisition with backoff for completion event delivery, so concurrent completions don't starve each other on the session file lock.