The four that rise
For four of the five, a higher figure is the worse one. Their levels must rise: the alerting level no higher than the acting one, and the acting no higher than the halting one.
The one that falls
The share of the venue's allowance already used runs the other way here — this stage treats a lower figure as the worse — so its levels must fall instead: alerting no lower than acting, acting no lower than halting.
A policy written for the four and applied to this one is therefore refused, with a complaint naming the measure.
The comparison that is not made
Each comparison names two levels and is made only where both are set.
So a policy giving an alerting level and a halting one, and leaving the acting level out, has those two never compared with each other — and a pair in the wrong order passes.
Related
- The order the health stage's checks stand in
- What the health policy refuses — the orderings the separate weigher insists on, over thresholds of its own
- What each check weighs
Source evidence
---
row: N57-02
baseline: 49aa659
created: 2026-11-08 18:12 UTC
modified: 2026-11-08 18:12 UTC
evidence:
- claim: "Four of the five measures are held to levels that rise: alerting no higher than acting, acting no higher than halting"
source: "`nexus/core/validator/health_stage.py` 111-115, 116-120, 121-125 and 131-135 into 323-328, and 337-342"
- claim: "The share of the allowance used is held to levels that fall instead, a lower figure being the worse for it"
source: "`nexus/core/validator/health_stage.py` 126-130 into 329-334, and 343-348"
- claim: "Each comparison is made only where both the levels it names are set"
source: "`nexus/core/validator/health_stage.py` 322, and 336"
- claim: "So a middle level left out leaves the outer two never compared with each other"
source: "`nexus/core/validator/health_stage.py` 322 and 336, neither comparing the outer pair"
---