Users report that a scheduling agent occasionally books the wrong meeting room, yet its final messages name the correct room. Logs capture only the user's request and the final response. Which instrumentation change best enables root-cause analysis?

1. A Log the final response plus a rationale the model writes at the end describing which steps it took and why it chose them.
2. B Raise log verbosity on the booking API so every request it receives, with full parameters, is captured for later review.
3. C Store full transcripts only for runs that end in an error, and sample a small fraction of the remaining runs to keep storage costs under control.
4. D Record a trace per run with a span for each model and tool call: inputs, exact emitted arguments, raw results, latency, tokens.