{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp09_archetype_breadth150_20260804","research_id":"eoa_inverse_innovation_exp09_light_prior_art_20260804","cell_id":"load_balancing__computer_science","search_lanes":{"direct_problem_and_intervention":{"queries":["code review reviewer recommendation workload load balancing pull requests research","automatic code reviewer assignment workload expertise availability monorepo"],"source_ids":["SRC1","SRC2","SRC3","SRC4"],"no_result_note":null},"synonyms_and_historical_terms":{"queries":["workload-aware code reviewer recommendation","reviewer recommendation workload code review algorithm expertise","code review assignment expertise workload multi objective reviewer recommendation primary study"],"source_ids":["SRC1","SRC4"],"no_result_note":null},"products_practices_and_standards":{"queries":["GitHub CODEOWNERS reviewer assignment load balance round robin least busy","CODEOWNERS automatic reviewer assignment workload availability conflicts separation duties code review","GitLab automatic reviewer assignment approval rules workload availability"],"source_ids":["SRC2","SRC3"],"no_result_note":null},"component_combination":{"queries":["code reviewer recommendation expertise active workload availability","code owner approval rules eligible approvers review workload","reviewer recommendation workload skew historical simulation"],"source_ids":["SRC1","SRC2","SRC3","SRC4"],"no_result_note":null}},"sources":[{"source_id":"SRC1","title":"Improving Code Reviewer Recommendation: Accuracy, Latency, Workload, and Bystanders","publisher":"Meta researchers; arXiv","url":"https://arxiv.org/abs/2312.17169","source_type":"PRIMARY_RESEARCH","claims_supported":["Meta operated a production reviewer recommender and developed a workload-balanced version.","Reviewer workload was unevenly distributed.","A randomized production test of workload-balanced recommendations covered 28,000 diff authors and produced mixed results, demonstrating both feasibility and unresolved outcome uncertainty."]},{"source_id":"SRC2","title":"Managing code review settings for your team","publisher":"GitHub","url":"https://docs.github.com/en/organizations/organizing-members-into-teams/managing-code-review-settings-for-your-team","source_type":"OFFICIAL_GUIDANCE","claims_supported":["GitHub can automatically replace a requested team with selected individual reviewers.","Its load-balance algorithm considers recent request counts and outstanding reviews.","Reviewers marked Busy are excluded.","Required-code-owner branch protection can preserve the team approval requirement while individuals are assigned."]},{"source_id":"SRC3","title":"Automatic reviewer assignment","publisher":"GitLab","url":"https://docs.gitlab.com/user/project/merge_requests/reviews/automatic_reviewer_assignment/","source_type":"FIRST_PARTY_PRODUCT","claims_supported":["GitLab documents automatic assignment from CODEOWNERS.","Its testing-stage Duo strategy reads each approval rule, selects the minimum reviewers needed to satisfy it, and explains the choice.","Selection among eligible approvers considers availability, open-review workload, local time, and recent activity.","The advanced strategy is explicitly available for testing but not ready for production use."]},{"source_id":"SRC4","title":"Factoring Expertise, Workload, and Turnover into Code Review Recommendation","publisher":"Concordia University researchers; arXiv","url":"https://arxiv.org/abs/2312.17236","source_type":"PRIMARY_RESEARCH","claims_supported":["Historical analyses found highly concentrated review workload, with roughly 20% of reviewers performing 80% of reviews across studied projects.","SofiaWL recommends knowledgeable reviewers with low active-review workload.","Historical replay and simulation are established evaluation methods for workload-aware reviewer recommendation.","The reported approach increased expertise, reduced workload concentration, and reduced files at risk in the studied projects, while also documenting tradeoffs among objectives."]}],"problem_evidence":{"status":"PARTLY_SUPPORTED","finding":"The core phenomenon is visible: empirical research reports strongly concentrated review workloads, and major code-review platforms expose load-aware assignment and availability controls. The retained public evidence does not isolate the proposal's narrower monorepository condition—static ownership or habitual selection causing queues despite verified, policy-equivalent idle reviewers—so that local causal and substitutability claim remains unverified.","source_ids":["SRC1","SRC2","SRC3","SRC4"]},"closest_prior_art":[{"name":"GitLab Duo automatic reviewer-assignment strategy","source_ids":["SRC3"],"overlap":"Reconstructs eligible approvers from required approval rules, then chooses among them using availability and open-review workload and exposes an explanatory note. This substantially matches compatibility-first, health-aware, capacity-aware routing.","remaining_difference":"The documentation does not specify estimated change effort, recent-burden fairness bounds, conflict and protected-code checks beyond approval eligibility, continuity and cooldown behavior, or a read-only historical shadow replay; the feature is testing-stage."},{"name":"GitHub team code-review auto assignment with load balancing","source_ids":["SRC2"],"overlap":"Automatically selects individual reviewers from a requested team using recent assignments, outstanding reviews, and Busy status while retaining required-code-owner protection.","remaining_difference":"It is team-scoped and count-based; the documentation does not describe per-change expertise reconstruction, effort-weighted capacity, separation-of-duty or conflict filters, continuity cooldowns, or the proposed audit protocol."},{"name":"Meta RevRecWL","source_ids":["SRC1"],"overlap":"A production-scale reviewer recommender was modified to balance reviewer workload and evaluated by randomized experiment, directly colliding with the central destination-selection mechanism.","remaining_difference":"The retained source does not establish the proposal's full authorization-first constraint set, estimated-effort weights, cooldown and continuity invariants, or repository-local shadow audit; its mixed trial results leave outcome improvement open."},{"name":"SofiaWL and workload-aware WhoDo lineage","source_ids":["SRC4"],"overlap":"Combines reviewer expertise with low active-review workload, documents concentrated review work, and evaluates recommendations through historical simulation, closely matching both mechanism and proposed first test.","remaining_difference":"Its objectives include knowledge distribution and turnover, and the retained source does not document enforcement of repository approval, access, conflict, independence, protected-code, continuity, or cooldown constraints."}],"prior_art_disposition":"ESTABLISHED_PRACTICE","contrastive_claim_remaining":"Relative to existing load-aware reviewer recommenders, a repository-specific router that first reproduces every approval, access, conflict, independence, and protected-change constraint and then uses effort-weighted active workload, recent-burden bounds, continuity, and cooldown rules may reduce eligible queue-age and assignment skew without any invariant violation. This is a narrow implementation-and-outcome claim, not a claim to the general idea of workload-aware reviewer routing.","contrastive_claim_falsifier":"In a timestamp-ordered shadow replay against the strongest available baseline, the extra constraint reconstruction and effort, burden, continuity, and cooldown features either never materially change destinations, fail to improve queue-age or capacity-normalized skew, increase unassigned work or flapping, or recommend even one ineligible reviewer.","gates":{"adequate_source_search":{"status":"PASS","rationale":"Direct, synonym, product/practice, and component-combination searches found and opened two primary research sources and two current first-party or official product sources from four publisher contexts.","source_ids":["SRC1","SRC2","SRC3","SRC4"]},"supported_problem":{"status":"PASS","rationale":"The general workload-concentration problem is empirically supported, although its monorepository-specific cause and the existence of truly interchangeable idle reviewers require local verification.","source_ids":["SRC1","SRC4"]},"distinct_testable_claim":{"status":"PASS","rationale":"A narrow contrast remains around complete policy reconstruction plus effort-weighted load, burden bounds, continuity, and cooldowns, with measurable destination, queue-age, skew, flapping, and eligibility outcomes. The broad routing concept itself is not distinct.","source_ids":["SRC1","SRC2","SRC3","SRC4"]},"bounded_next_test":{"status":"PASS","rationale":"A 14-day, one-repository, read-only timestamp replay with a 30-request manual audit is bounded and compares observable outcomes without changing live assignments. Because research warns that historical and live experimental results may diverge, it is an initial feasibility and safety test rather than proof of operational value.","source_ids":["SRC1","SRC4"]},"no_obvious_safety_or_authority_stop":{"status":"PASS","rationale":"Shadow-only recommendations preserve existing approval authority and avoid live invitations, permission changes, approvals, or merges. The test should stop on failed eligibility reconstruction or protected-data exposure; no source reveals an unavoidable authority or safety prohibition.","source_ids":["SRC2","SRC3"]}},"screen_survival":false,"world_novelty_boundary":"This bounded public search establishes substantial, implemented prior art for workload- and availability-aware code-review routing, including selection within approval-rule-eligible pools. It does not establish exhaustive world novelty, patentability, market size, expert acceptance, realized value, or whether the narrower constraint-complete and effort-weighted variant exists in undisclosed internal systems."}