Claude Code’s background-agent summaries have drawn scrutiny because the feature reportedly resends conversation history roughly every 30 seconds to generate a short status label. One company’s OpenTelemetry analysis attributed 19.7% of API requests and 16.4% of estimated spending to it. A developer representative instead cited about 2% across overall telemetry and about 6% for terminal users. The difference highlights workload and model dependence, including how many background agents run and for how long. The report did not identify an official switch to disable the feature, and individual measurements should not be presented as a universal waste rate.