Parallel tool results

An agent that queries three services has started making tool calls one at a time instead of in parallel, roughly a week after this harness shipped. Nothing errors, and every result reaches Claude. Find why the parallelism decayed.

tool_uses = [b for b in response.content if b.type == "tool_use"]
messages.append({"role": "assistant", "content": response.content})

# Execute concurrently, then hand each result back as it lands.
for block in tool_uses:
    result = run_tool(block.name, block.input)
    messages.append({
        "role": "user",
        "content": [
            {"type": "tool_result", "tool_use_id": block.id, "content": result}
        ],
    })