Extracted default-branch inclusion logic from CommitCompleteJob into a new CommitDefaultBranchConfirmer job (app/jobs/commit_default_branch_confirmer.rb) and added denormalization hooks in CommitSite to invoke it when a commit site has kin on the default branch. This adds a redundant call site to improve reliability of marking commits as present in the main/default branch.

Adds "Test Delta Cohorts" and "Copy/Paste Cohorts" to the AI Cohort stats lineup and refreshes the chart color palette. Fixes palette retrieval in area-chart-segment-colors.js so AreaChart no longer depends on a missing segment prop (derives palette from chartDataPath), preventing incorrect or missing series colors. Result: new cohort options appear and charts render with consistent colors.

Renamed the UI/stat label from "Commit count per AI Cohort" to "Commit Count per Count in AI Cohort" across the stats view and related presenters/helpers (notably app/views/shared/resource_actions/ai_cohorts_stats.html.haml, presenters and defines). These find-and-replace updates standardize phrasing so cohort commit metrics display a clearer, corrected label in the UI.

Rework the redirect decision logic in the repo import flow so that, when there is no evidence that a user has completed the GK redirect, the GK redirect path takes precedence over other redirect types (except the usage/profile path). This change touches app/lib/controller_concerns/repo_import_concern.rb and app/lib/git_kraken.rb, adjusting the priority rules and how completion evidence is evaluated. Benefit: more predictable, correct redirects when GK completion is uncertain, reducing incorre...

Consolidates the GitKraken redirect logic from UserUsageProfilesController into a new app/lib/controller_concerns/kraken_controller_concern.rb as KrakenControllerConcern. The moved methods, including notify_gitkraken_and_redirect and confirm_provider_import_with_gitkraken, are now accessible via import or the user usage profile, improving separation of concerns and testability. This modularization enhances reuse and maintainability of the GK redirect behavior.

Moved CaptureCommitterDetail out of the committer callback and into a post-work method invoked by ProcessCommit and ClosedPR, and added Committer#capture_provider_id_if_applicable to fetch missing GitHub provider IDs via GraphQL. Tests for fetching missing remote IDs now pass. May later call the same post-work after PR review creation. Impact: more reliable retrieval of missing GitHub IDs for accurate commit/PR attribution.

Return a new default_branch_processed_back_through column on import_progress so callers can determine how far back the default-branch import has progressed. Changes touch RepoConcerns::ImportProgress and RepoImportProgress and update the API reference partial/helper to expose the field. Improves accuracy of import status reporting and simplifies progress/resume logic.

Verify that commits participating in a closed pull request are on the repository default branch. Adds checks in ProcessClosedPullRequest (ensure_merged_commits_on_default_branch), updates CommitSite.kin_on_default_branch?, and expands ProcessClosedPullRequest tests. Prevents processing or attributing commits that were merged outside the default branch, improving accuracy of merged-commit handling.

Add provider_email to the developer_list JSON and update lookup logic to fall back to matching developers by email when the supplied git_provider_id does not match any known developer. Changes touch developer_list serialization and API helper/ controller logic so responses include provider_email and developers are resolved more reliably, reducing failed or incomplete developer records in API responses.

Adds a new teammates_review_minutes committer_stat and LlmTeamCohortStat cohort-building to measure teammates' review minutes and assess whether AI increases review burden. Extracts start_date and date_type_em validation from StatCacheConcerns into a standalone concern since AI stat models don't use digests. Deployed to production Oct 22, 2025 — enables team-level review-time metrics for monitoring AI impact.

Added a unit test (vendor/gems/llm_metrics/test/unit/github_copilot_different_entity_test.rb) and a VCR cassette to validate processing of Copilot stats from Bitbucket, GitLab and Azure organizations, plus minor updates to app/models/provider_resource_mirror.rb. This targets IN-1 (list developers endpoint returned duplicates) by exercising different-entity handling and improves reliability of stats aggregation in production.

Added a test to verify that Copilot stats can be processed for Bitbucket, Gitlab, and Azure organizations. The test exercises the processing path end-to-end and passes in CI, giving greater confidence against regressions. No specific application files are indicated as changed, so the update appears to focus on test coverage. Impact: improves reliability of Copilot stats processing across these providers and reduces risk of regression in production.