Race condition between large PreToolUse hook and cmd output.

- Copilot Chat Extension Version: Latest as of 2026-09-07
- VS Code Version: Latest as of 2026-09-07
- OS Version: macOS 26
- Feature (e.g. agent/edit/ask mode):
- Selected model (e.g. GPT 4.1, Claude 3.7 Sonnet): All

In the following setup:
- LLM makes bash cmd tool call generating large output (e.g. reading log)
- A custom PreToolUse hook script itself also produces large output

Both outputs get redirected to file `content.txt`, rather than inlined in conversation.
But there seems to be a race condition, and these writes step on each other.
In many cases, the write from hook output wins, and LLM agent will say "tool output polluted by hook" after reading `context.txt`.