What happened
PR #7956 was reviewed by the review agent at 05:17-05:24 UTC (workflow run 28919376383, APPROVED). At 08:40 UTC, Renovate rebased the branch to pick up base branch changes, producing a new commit SHA (c0f5058). This triggered a second review workflow (run 28929447407, 08:40-08:45 UTC) which re-ran the full review and reached the same APPROVED conclusion with no new findings. The semantic content of the digest changes was identical across both commits — the rebase only moved the base pointer forward.
What could go better
The second review consumed ~4.5 minutes of agent time and tokens with zero incremental value. For bot-authored dependency PRs where the rebase doesn't change the actual diff (only the base commit moves forward), a full re-review is wasteful. This adds to evidence from PR #7924 (issue #3401, 3 redundant runs) and PR #816 (the original #1355 filing). Confidence is high.
Proposed change
This is evidence for existing issue #1355. The proposed change remains: implement a diff-aware skip mechanism in the review dispatch that compares the semantic diff of the new commit against the previously-reviewed commit. When the diff is identical (only the base SHA changed), skip the review and carry forward the previous approval. This should apply specifically to bot-authored PRs to minimize risk.
Validation criteria
Renovate rebases on unchanged diffs should not trigger new review runs. Success metric: zero redundant review runs from rebases on bot dependency PRs over 30 days, with no cases where a meaningful diff change was missed.
Generated by retro agent from konflux-ci/konflux-ci#7956
What happened
PR #7956 was reviewed by the review agent at 05:17-05:24 UTC (workflow run 28919376383, APPROVED). At 08:40 UTC, Renovate rebased the branch to pick up base branch changes, producing a new commit SHA (
c0f5058). This triggered a second review workflow (run 28929447407, 08:40-08:45 UTC) which re-ran the full review and reached the same APPROVED conclusion with no new findings. The semantic content of the digest changes was identical across both commits — the rebase only moved the base pointer forward.What could go better
The second review consumed ~4.5 minutes of agent time and tokens with zero incremental value. For bot-authored dependency PRs where the rebase doesn't change the actual diff (only the base commit moves forward), a full re-review is wasteful. This adds to evidence from PR #7924 (issue #3401, 3 redundant runs) and PR #816 (the original #1355 filing). Confidence is high.
Proposed change
This is evidence for existing issue #1355. The proposed change remains: implement a diff-aware skip mechanism in the review dispatch that compares the semantic diff of the new commit against the previously-reviewed commit. When the diff is identical (only the base SHA changed), skip the review and carry forward the previous approval. This should apply specifically to bot-authored PRs to minimize risk.
Validation criteria
Renovate rebases on unchanged diffs should not trigger new review runs. Success metric: zero redundant review runs from rebases on bot dependency PRs over 30 days, with no cases where a meaningful diff change was missed.
Generated by retro agent from konflux-ci/konflux-ci#7956