馃敶 HIGH SEV 路 Inference Consumer Starved (no records from device-events)
馃敟 Prod consumer is reading < 0.5 records/s from device-events
<|馃搱 graph> 路 <|馃搳 service dashboard> 路
馃毃 alarm definition 路
馃敃 silence
HIGH SEV - permanent loss, and the worst customer outcome in the set. No events means foreground detection stops entirely: screen time stops accruing, limits never trigger, and every kid on the platform silently has unlimited access. Because the events were never produced there is NO backfill - that time is permanently unaccounted, unlike a consumer-side outage which replays from the log. This is the 2026-07-14 shape that ran ~1.5 days undetected (CLE-456). Prod baseline 14.6/s with a diurnal floor of 4.7/s, so 0.5 is ~10x margin. This rule is the AGGREGATE catch-all; for which producer died see 'Inference Proxy Producer Silent' / 'Inference API Producer Silent'.
Grafana vmain-b1e6358