Latency and Data-Staleness Indicators
Operators must always know how fresh what they are looking at is; the interface shows data age and freezes rather than extrapolating when a feed stalls.
A number without an age is a hazard
Every value on an operator surface has an age — the time since it was actually measured or computed — and a stale value that looks live is one of the most dangerous things an interface can show. Kronos L6 makes freshness explicit. Each panel knows the timestamp of its underlying data (from the PTP-disciplined L2 fabric) and surfaces staleness the moment it exceeds what that panel's decisions tolerate.
Freshness budgets per surface
| Surface | Tolerated age before staleness shown |
|---|---|
| 3D twin overlay / shape | tens of milliseconds |
| Fast KPIs (current, plug density) | tens to hundreds of ms |
| Slow process KPIs (cryo, vacuum trend) | seconds |
| Mobile / off-console views | seconds, with connectivity note |
When a feed stalls, the panel freezes its last-known value and marks it stale rather than extrapolating a plausible-looking continuation — because a confident-looking guess is worse than an honest gap. The staleness badge shows how old the data is and, where relevant, why (source down, network gap, diagnostic dropped). On mobile, a connectivity gap always shows as staleness, never as a silently frozen dashboard.
Latency of the pipeline itself
Beyond per-value age, the interface exposes end-to-end latency of the observability path so the crew knows how far behind reality the picture inherently is. This is distinct from the hard-real-time control latency, which lives in L1; the L6 latency is the human-observation lag. Showing it keeps operators calibrated: a 50 ms-old overlay is fine for supervision but is not a substitute for the microsecond loops L1 runs.
Staleness ties directly into confidence and provenance — an imputed or stale input lowers an estimate's confidence — see provenance display. It also protects the shared video wall and mobile views from showing frozen truth.