What happened
PR #22 was authored by red-hat-konflux[bot], created 2025-11-05, and went through dozens of force-push/rebase cycles over 8 months before being autoclosed by Renovate/MintMaker on 2026-07-08T18:40:02Z without being merged (merged_at is null, title includes '- autoclosed' suffix). The retro agent was dispatched at 2026-07-08T18:40:55Z, less than 1 minute after closure, triggering a full retro analysis via workflow run 28966934852.
What could go better
Running a full retro analysis on a bot-authored dependency PR that was autoclosed without merge has limited value. The code changes were never integrated, so there is no production impact to analyze. While the false approval finding (documented separately) is worth capturing, the dispatch logic should detect this class of PR and either skip the retro entirely or use a lightweight fast-path.
This pattern is already well-documented: #3700 (autoclosed bot PR in project-controller), #3698 (autoclosed bot PR in crossplane-components), #3618 (autoclosed bot PR in caching). The companion PR #121 (genproto/googleapis/api) was also autoclosed on the same date, likely triggering another retro.
Confidence: high that this retro dispatch adds minimal value relative to its cost.
Proposed change
This is additional evidence for fullsend-ai/fullsend#3317 (skip or use lightweight retro for bot-authored PRs) and #3647 (skip retro for automerged bot dependency PRs). The dispatch-level filter proposed in those issues should be extended to also cover autoclosed (not just automerged) bot PRs: when a bot-authored dependency PR is closed without merge, skip the retro stage. No new proposal is needed.
Validation criteria
After implementing #3317/#3647, autoclosed (never-merged) Renovate/MintMaker PRs should not trigger full retro agent runs. Measure: zero retro dispatches for autoclosed bot dependency PRs over a 14-day window across the konflux-ci org.
Generated by retro agent from konflux-ci/devlake#22
What happened
PR #22 was authored by red-hat-konflux[bot], created 2025-11-05, and went through dozens of force-push/rebase cycles over 8 months before being autoclosed by Renovate/MintMaker on 2026-07-08T18:40:02Z without being merged (merged_at is null, title includes '- autoclosed' suffix). The retro agent was dispatched at 2026-07-08T18:40:55Z, less than 1 minute after closure, triggering a full retro analysis via workflow run 28966934852.
What could go better
Running a full retro analysis on a bot-authored dependency PR that was autoclosed without merge has limited value. The code changes were never integrated, so there is no production impact to analyze. While the false approval finding (documented separately) is worth capturing, the dispatch logic should detect this class of PR and either skip the retro entirely or use a lightweight fast-path.
This pattern is already well-documented: #3700 (autoclosed bot PR in project-controller), #3698 (autoclosed bot PR in crossplane-components), #3618 (autoclosed bot PR in caching). The companion PR #121 (genproto/googleapis/api) was also autoclosed on the same date, likely triggering another retro.
Confidence: high that this retro dispatch adds minimal value relative to its cost.
Proposed change
This is additional evidence for fullsend-ai/fullsend#3317 (skip or use lightweight retro for bot-authored PRs) and #3647 (skip retro for automerged bot dependency PRs). The dispatch-level filter proposed in those issues should be extended to also cover autoclosed (not just automerged) bot PRs: when a bot-authored dependency PR is closed without merge, skip the retro stage. No new proposal is needed.
Validation criteria
After implementing #3317/#3647, autoclosed (never-merged) Renovate/MintMaker PRs should not trigger full retro agent runs. Measure: zero retro dispatches for autoclosed bot dependency PRs over a 14-day window across the konflux-ci org.
Generated by retro agent from konflux-ci/devlake#22