Context
The DeepReport 2026-05-17 (#32887) flagged: "Deployment Monitor: 100 runs/day at 8% success rate = ~92 wasted invocations/day (zombie pattern)". The Copilot Session Insights 2026-05-18 (#32986) reinforces this: 37 of 50 (74%) recent sessions ended in action_required with zero duration — workflows triggering on events they shouldn't, or blocked by a required check by default.
Hypothesis
The workflow fires on a broad event trigger but most invocations have no actionable deployment to monitor — so they correctly no-op but still consume runner minutes, observability budget, and clutter the daily-failure dashboards. The 92 daily wasted runs are the largest single source of noise in the Copilot session analysis.
Why it matters
- Action minutes: 92 wasted runs/day × ~5 min/run ≈ 7.7 hours/day of runner time
- Observability noise: Every wasted run contributes to the OTLP 401 storm and dashboard clutter
- Daily-news false-positive load: These runs surface as
action_required and crowd out signal
Proposed investigation
- Read
.github/workflows/deployment-incident-monitor.md (also referenced in #33052)
- Identify which event(s) trigger it (likely
workflow_run or push)
- Add a
pre_activation gate that exits early when there is no incident state to inspect (e.g. only run when a recent workflow_run conclusion is failure for the monitored workflows)
- Alternatively, narrow the trigger to a scheduled cadence + manual dispatch
Acceptance criteria
- Success rate climbs above 50% (≥45 productive runs out of ≤90)
- OR run-count drops by ≥80% with the same productive output
Suggested agent
Workflow Normalizer, or CI Optimization Coach — both have prior context on workflow gating.
Estimated effort
Small-medium: workflow markdown edit + a pre_activation JS gate + verify next 24h of runs.
Source
DeepReport 2026-05-17 #32887; Copilot Session Insights 2026-05-18 #32986.
Generated by 🔬 DeepReport - Intelligence Gathering Agent · ● 19.5M · ◷
Context
The DeepReport 2026-05-17 (#32887) flagged: "Deployment Monitor: 100 runs/day at 8% success rate = ~92 wasted invocations/day (zombie pattern)". The Copilot Session Insights 2026-05-18 (#32986) reinforces this: 37 of 50 (74%) recent sessions ended in
action_requiredwith zero duration — workflows triggering on events they shouldn't, or blocked by a required check by default.Hypothesis
The workflow fires on a broad event trigger but most invocations have no actionable deployment to monitor — so they correctly no-op but still consume runner minutes, observability budget, and clutter the daily-failure dashboards. The 92 daily wasted runs are the largest single source of noise in the Copilot session analysis.
Why it matters
action_requiredand crowd out signalProposed investigation
.github/workflows/deployment-incident-monitor.md(also referenced in #33052)workflow_runorpush)pre_activationgate that exits early when there is no incident state to inspect (e.g. only run when a recent workflow_run conclusion isfailurefor the monitored workflows)Acceptance criteria
Suggested agent
Workflow Normalizer, or CI Optimization Coach — both have prior context on workflow gating.
Estimated effort
Small-medium: workflow markdown edit + a
pre_activationJS gate + verify next 24h of runs.Source
DeepReport 2026-05-17 #32887; Copilot Session Insights 2026-05-18 #32986.