Measure around-the-clock work with event coverage and redundancy, not open sessions

Around-the-clock activity should be measured from bounded event-time intervals, with weekends and distinct-owner redundancy reported separately. An open session or a healthy HTTP response is insufficient: check the underlying source watermark before interpreting zero activity. Keep observed activity, continuous attention, useful completed work, and reliable operational coverage as separate claims. This method applies to coding-agent and scheduled-work telemetry; it cannot establish uninstrumented customer operations or human staffing.

A practical audit counts occupied 15-minute bins in a fixed seven-day window, excludes replay/test identities, retains timestamp provenance, and tests sensitivity to removing one dominant owner. Union sources for evidence coverage without adding duplicate volumes; APIs backed by the same store are not independent witnesses. A collector gap remains unknown, not proof of inactivity. Even 100% hourly occupancy can conceal shorter outages.

Evidence/example — 2026-09-16: an authorized E3 telemetry audit found a presence endpoint with a current response-generation timestamp but a months-old source watermark, alongside fresh coding events elsewhere. Narrow raw message reads also filled intervals absent from the main telemetry feed, demonstrating why a single-source gap should be qualified. Sources: Cosmos heartbeat__who_is_active / presence_history, Sherlock diagnostics / coverage, and bounded TimeTracker event reads. Service implementations: https://github.com/e3-solutions/e3-heartbeat and https://github.com/e3-solutions/sherlock . No personnel records, raw logs or private task content are included. This observation supports the cited general principle that partial traces support bounded findings, not complete-success claims; it does not prove the cause of the stale projection.