EPS Dashboard

Overview / Motivation

Auto-filed by /daily 2026-08-02 (route 2: behavior/logic change -> independent review) from the nightly problem sweep (miner6, session a0400dd4, task #1345).

Goal

The inline payload lint gate does not false-block payloads because the shared VM happened to be at load 30 when the mapped tests ran.

Workflow gap

Bug observed: A certified inline-gate run FAILed exit 3 under VM load1~31 (1 failed + 3 error mapped tests, 8m34s pytest) and PASSed clean ~35 min later once load dropped -- the session improvised a load-drop Monitor watch and retried (session a0400dd4, 17:08-18:25Z, task #1345). unverified hypothesis -- verify at plan time: the failures were load-flake (timeout-sensitive tests), inferred from the clean retry at low load.
Why it is a workflow gap: The gate's mapped-pytest leg is timing-sensitive; fleet-busy hours make gate verdicts load-dependent, and each false BLOCK costs a manual retry loop the session must improvise.
Confidence (emitter): low
verified-at-filing: grep -c -iE 'getloadavg|loadavg|load1' scripts/inline_lint_gate.py -> 0 (no load awareness).

Proposed change (refine in planning)

check os.getloadavg() before the mapped-pytest leg; above a threshold (~20 on this 32-core VM) either wait bounded for load to drop or return a distinct INCONCLUSIVE-under-load verdict, so a loaded fleet doesn't convert to false BLOCKs.

Scope / surfaces

Primary target: scripts/inline_lint_gate.py

Constraints / invariants

Workflow-surface rules apply; scripts/workflow_lint.py --check-asks passes; ruff on touched files passes.
This session runs under EPM_WORKFLOW_FIX_SESSION=1 (Provenance workflow_fix_target: line) -- it MUST NOT auto-route its own subagents' workflow-fix candidates (recursion guard).

Provenance

fingerprint: 517ebd1cb2be
workflow_fix_target: scripts/inline_lint_gate.py