Recap & Practice

Why is it wasteful for three independent API calls to run strictly one after another, each with its own separate wait?

Implement run_tools: start log_run_start(log) in the background immediately using create_task, run every call_tool(...) for tool_specs concurrently via asyncio.gather with return_exceptions=True, await the logging task before returning, and split the gather results into (successes, failures).