Summary
Overall status: Mixed — auto-labelling pipeline is operational but the correction pipeline is fully stalled.
The Label Discussions workflow ran successfully every day last week (7/7 runs), indicating active discussion processing. However, the Labelling Correction Collector has zero runs in the last 30 days, no new correction signals have been created since 2026-04-18, and 333 of 335 correction signals remain open. The backlog is not shrinking.
Key Metrics
| Metric |
Last 7 Days |
Prev 7 Days |
| Discussions reviewed |
N/A — no parsed summary issues |
N/A |
| Label changes applied |
N/A — no parsed summary issues |
N/A |
| Change rate (changed / reviewed) |
Not calculable |
Not calculable |
Label Discussions runs (success) |
7 / 7 |
7 |
Labelling Correction Collector runs |
0 |
0 |
Labelling Correction Feedback runs (skipped) |
6 / 6 |
7 |
| Open correction signals |
333 |
— |
| New correction signals last 7 days |
0 |
0 |
| New correction signals last 30 days |
0 |
0 |
⚠️ No daily summary issues were available to parse. Reviewed and changed counts are unavailable; auto-labelling throughput cannot be confirmed beyond "runs succeeded."
⚠️ The Labelling Correction Collector has not run in the last 30 days. This is the primary signal-ingestion workflow; its absence explains why no new correction signals have appeared.
Correction Pressure
All 333 open correction signals were created between 2026-03-31 and 2026-04-18 (Batch 02). No new signals have been ingested since. The backlog represents 851 total correction events across 333 discussions.
The highest-pressure categories by open-signal count:
| Category |
Open Signals |
| Copilot Conversations |
110 |
| GitHub Education |
71 |
| Other Feature Feedback, Questions & Ideas |
52 |
| Apps, API and Webhooks |
21 |
| Mobile |
14 |
| Discussions |
13 |
| Actions |
10 |
The most-corrected labels (by latest_label on open signals):
| Label |
Signals |
| Copilot |
56 |
| Copilot in GitHub |
42 |
| GitHub Education |
38 |
| bug |
37 |
| Other Features and Feedback |
20 |
Event type split: 197 labeled vs 136 unlabeled events — meaning the system is adding wrong labels slightly more often than it is failing to add correct ones.
100 signals have correction_count ≥ 2, indicating repeated human intervention on the same discussions. The top 5 most-corrected signals:
| Signal |
Discussion |
correction_count |
Category |
#114 |
#147 |
19 |
Other Feature Feedback |
#69 |
#185 |
17 |
GitHub Education |
#63 |
#195 |
17 |
Other Feature Feedback |
#55 |
#189 |
17 |
GitHub Education |
#51 |
#186 |
17 |
Other Feature Feedback |
All 333 open signals are concentrated in 4 open parent intake issues (Batch 02: #45, #46, #49, #50). No signals are spread outside these batches.
Open correction signal breakdown (top 15 by correction_count)
| Issue |
Discussion |
correction_count |
Category |
Latest Label |
#114 |
#147 |
19 |
Other Feature Feedback, Questions & Ideas |
Code Search and Navigation |
#69 |
#185 |
17 |
GitHub Education |
question |
#63 |
#195 |
17 |
Other Feature Feedback, Questions & Ideas |
Profile |
#55 |
#189 |
17 |
GitHub Education |
GitHub Education Benefits |
#51 |
#186 |
17 |
Other Feature Feedback, Questions & Ideas |
Code Search and Navigation |
#109 |
#151 |
16 |
Copilot Conversations |
Copilot |
#75 |
#175 |
15 |
GitHub Education |
GitHub Education |
#118 |
#148 |
14 |
Apps, API and Webhooks |
Apps API and Webhooks |
#77 |
#170 |
14 |
Copilot Conversations |
Copilot |
#73 |
#184 |
14 |
Other Feature Feedback, Questions & Ideas |
Copilot |
#108 |
#152 |
12 |
Copilot Conversations |
Copilot |
#74 |
#176 |
12 |
Copilot Conversations |
Copilot |
#57 |
#187 |
11 |
GitHub Education |
GitHub Education |
#79 |
#167 |
11 |
Other Feature Feedback, Questions & Ideas |
Profile |
#83 |
#163 |
11 |
Copilot Conversations |
Copilot |
Open Instruction Debt
The correction backlog is steady (not shrinking). With the Correction Collector not running, the queue is frozen rather than growing or resolving.
- Oldest open correction signal:
#47, created 2026-04-08 — 44 days old
- Open parent intake issues: 4 (all Batch 02; Batch 01 is the only closed one)
- Closed signals: 2 of 335 (< 1%)
The backlog does not look actionable in its current state: no workflow is collecting new corrections and no process is closing resolved signals. The Correction Feedback runs are all skipped, which is consistent with no new signals arriving as triggers. Until the Collector resumes, the backlog will remain frozen and diverge further from the live labelling state.
Recommendations
-
Investigate and restore the Labelling Correction Collector workflow. It has not run in at least 30 days. This is the root cause of the frozen correction pipeline. Check for disabled schedules, failed permissions, or quota exhaustion.
-
Review .github/instructions/community-discussion-labeling.md for the three highest-pressure clusters: Copilot/Copilot-in-GitHub label confusion, GitHub Education signal patterns, and the bug vs. question vs. Other Features and Feedback ambiguity. The 100 signals with correction_count ≥ 2 represent confirmed instruction gaps.
-
Triage the 4 open Batch 02 parent intake issues (#45, #46, #49, #50). Determine whether they represent distinct rule-update opportunities or can be consolidated. The signals under them predate the current labelling instruction version and may partly be stale.
-
Enable daily summary issue publishing for Label Discussions runs (or verify the parsing pipeline). Without readable summary issues, throughput metrics (reviewed count, change rate) are completely invisible, making health reporting a lagging indicator.
References
Recent Label Discussions workflow runs (last 7 days)
- §79 — 2026-05-22 — success
- §78 — 2026-05-21 — success
- §77 — 2026-05-20 — success
- §76 — 2026-05-19 — success
- §75 — 2026-05-18 — success
- §74 — 2026-05-17 — success
- §73 — 2026-05-16 — success
Recent Labelling Correction Feedback runs (last 7 days, all skipped)
- §417 — 2026-05-22 — skipped
- §416 — 2026-05-21 — skipped
- §415 — 2026-05-20 — skipped
- §414 — 2026-05-18 — skipped
- §413 — 2026-05-17 — skipped
- §412 — 2026-05-16 — skipped
§79 · §78 · §77
Generated automatically by the Labelling Health workflow on 2026-05-23T03:34:00Z.
Generated by Labelling Health Report · ● 384.8K · ◷
Summary
Overall status: Mixed — auto-labelling pipeline is operational but the correction pipeline is fully stalled.
The
Label Discussionsworkflow ran successfully every day last week (7/7 runs), indicating active discussion processing. However, theLabelling Correction Collectorhas zero runs in the last 30 days, no new correction signals have been created since 2026-04-18, and 333 of 335 correction signals remain open. The backlog is not shrinking.Key Metrics
Label Discussionsruns (success)Labelling Correction CollectorrunsLabelling Correction Feedbackruns (skipped)Correction Pressure
All 333 open correction signals were created between 2026-03-31 and 2026-04-18 (Batch 02). No new signals have been ingested since. The backlog represents 851 total correction events across 333 discussions.
The highest-pressure categories by open-signal count:
The most-corrected labels (by
latest_labelon open signals):Event type split: 197
labeledvs 136unlabeledevents — meaning the system is adding wrong labels slightly more often than it is failing to add correct ones.100 signals have correction_count ≥ 2, indicating repeated human intervention on the same discussions. The top 5 most-corrected signals:
#114#147#69#185#63#195#55#189#51#186All 333 open signals are concentrated in 4 open parent intake issues (Batch 02:
#45,#46,#49,#50). No signals are spread outside these batches.Open correction signal breakdown (top 15 by correction_count)
#114#147#69#185#63#195#55#189#51#186#109#151#75#175#118#148#77#170#73#184#108#152#74#176#57#187#79#167#83#163Open Instruction Debt
The correction backlog is steady (not shrinking). With the Correction Collector not running, the queue is frozen rather than growing or resolving.
#47, created 2026-04-08 — 44 days oldThe backlog does not look actionable in its current state: no workflow is collecting new corrections and no process is closing resolved signals. The Correction Feedback runs are all
skipped, which is consistent with no new signals arriving as triggers. Until the Collector resumes, the backlog will remain frozen and diverge further from the live labelling state.Recommendations
Investigate and restore the
Labelling Correction Collectorworkflow. It has not run in at least 30 days. This is the root cause of the frozen correction pipeline. Check for disabled schedules, failed permissions, or quota exhaustion.Review
.github/instructions/community-discussion-labeling.mdfor the three highest-pressure clusters: Copilot/Copilot-in-GitHub label confusion, GitHub Education signal patterns, and thebugvs.questionvs.Other Features and Feedbackambiguity. The 100 signals with correction_count ≥ 2 represent confirmed instruction gaps.Triage the 4 open Batch 02 parent intake issues (
#45,#46,#49,#50). Determine whether they represent distinct rule-update opportunities or can be consolidated. The signals under them predate the current labelling instruction version and may partly be stale.Enable daily summary issue publishing for
Label Discussionsruns (or verify the parsing pipeline). Without readable summary issues, throughput metrics (reviewed count, change rate) are completely invisible, making health reporting a lagging indicator.References
Recent Label Discussions workflow runs (last 7 days)
Recent Labelling Correction Feedback runs (last 7 days, all skipped)
§79 · §78 · §77
Generated automatically by the Labelling Health workflow on 2026-05-23T03:34:00Z.