{"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_shedding__computer_science","search_lanes":{"direct_problem_and_intervention":{"queries":["monorepo code intelligence cancel superseded queued analysis tasks overload","language server stale request cancellation background indexing queue overload"],"source_ids":["SRC1","SRC2","SRC4"],"no_result_note":null},"synonyms_and_historical_terms":{"queries":["incremental compiler obsolete work cancellation latest revision coalescing","language server content modified cancellation stale document request"],"source_ids":["SRC1","SRC2","SRC3"],"no_result_note":null},"products_practices_and_standards":{"queries":["clangd background index queue drop duplicate tasks priority","rust-analyzer cancel stale analysis on revision change","Jira indexing background queue user request priority","Language Server Protocol ContentModified cancelRequest"],"source_ids":["SRC1","SRC2","SRC3","SRC4"],"no_result_note":null},"component_combination":{"queries":["cancel superseded tasks queue coalescing reserve capacity hysteresis language server","latest-only work queue deadline expiration interactive background analysis"],"source_ids":["SRC1","SRC2","SRC4"],"no_result_note":null}},"sources":[{"source_id":"SRC1","title":"BackgroundQueue.cpp Source File","publisher":"LLVM Project","url":"https://clang.llvm.org/extra/doxygen/BackgroundQueue_8cpp_source.html","source_type":"FIRST_PARTY_PRODUCT","claims_supported":["clangd has an explicit background-index task queue with progress statistics and priority adjustment.","Its admission path returns without enqueueing a task when a keyed duplicate is detected, providing direct prior art for discarding selected background indexing work before execution.","The implementation discusses staleness limitations and reindexing as potentially poor use of CPU, but does not implement latest-revision eviction, overload activation, deadlines, protected capacity, or hysteretic recovery."]},{"source_id":"SRC2","title":"Architecture - rust-analyzer","publisher":"rust-analyzer Project","url":"https://rust-analyzer.github.io/book/contributing/architecture.html","source_type":"OFFICIAL_GUIDANCE","claims_supported":["rust-analyzer separates state-changing or typing-blocking work from requests processed in the background.","When an edit advances the global revision, ongoing analysis can be cancelled because its result is stale and it blocks input modification.","This establishes supersession-aware cancellation in code intelligence, but not an overload-triggered queue-shedding and recovery controller."]},{"source_id":"SRC3","title":"Language Server Protocol Specification 3.18","publisher":"Microsoft","url":"https://github.com/microsoft/language-server-protocol/blob/gh-pages/_specifications/lsp/3.18/specification.md","source_type":"OFFICIAL_STANDARD","claims_supported":["LSP defines $/cancelRequest, RequestCancelled, and ContentModified outcomes for invalidated work.","The specification advises clients to cancel results that are no longer useful and permits servers to return ContentModified when an internal state change invalidates executing work.","It cautions servers against assuming that every older queued result is useless, supporting the proposal's need to establish supersession rather than infer it merely from a newer message."]},{"source_id":"SRC4","title":"Jira Indexing Queue","publisher":"Atlassian","url":"https://support.atlassian.com/jira/kb/jira-indexing-queue/","source_type":"FIRST_PARTY_PRODUCT","claims_supported":["Atlassian reports that bursts of background indexing could overflow a shared indexing queue, seriously affect user actions, and cause timeouts or inconsistent indexing.","Jira 8.16 separates user-request indexing into a primary queue and background work into a secondary queue consumed only when the primary queue is empty.","Jira records queue size and waiting-time statistics, but the documented remedy prioritizes classes rather than terminating superseded tasks or applying hysteretic overload recovery."]}],"problem_evidence":{"status":"PARTLY_SUPPORTED","finding":"The problem's component mechanisms are publicly visible: Atlassian documents background indexing overflowing a shared queue and harming foreground user actions, while rust-analyzer documents edits making ongoing code analysis stale and worth cancelling. The retained sources do not directly measure the combined scenario of superseded background work causing overload in a shared large-monorepo code-intelligence service.","source_ids":["SRC2","SRC4"]},"closest_prior_art":[{"name":"clangd BackgroundQueue duplicate rejection and priority boosting","source_ids":["SRC1"],"overlap":"A code-intelligence background-index queue selectively rejects keyed duplicate tasks before execution and supports priority changes, directly overlapping queue-boundary discard of redundant indexing work.","remaining_difference":"It rejects a newly submitted duplicate rather than establishing and retaining the newest revision, and it lacks overload thresholds, deadline expiration, protected interactive capacity, explicit supersession outcomes, and hysteretic re-entry."},{"name":"rust-analyzer revision-triggered cancellation","source_ids":["SRC2"],"overlap":"A new source revision cancels stale concurrent analysis so obsolete computation does not block updates.","remaining_difference":"Cancellation is driven by revision invalidation rather than measured overload and is not documented as a multi-tenant background-queue policy with capacity reservation, deadlines, audit categories, or staged recovery."},{"name":"Jira foreground/background indexing queues","source_ids":["SRC4"],"overlap":"Operational indexing practice separates user-triggered and background work after background bursts were observed to overflow a shared queue and harm user actions.","remaining_difference":"The secondary queue is deferred, not contracted by supersession-aware termination, and no dedicated worker reserve, usefulness deadline, or hysteretic re-admission policy is described."},{"name":"LSP cancellation and ContentModified semantics","source_ids":["SRC3"],"overlap":"The standard supplies cancellation and explicit invalidation outcomes for requests whose results are no longer useful or whose execution state has become invalid.","remaining_difference":"It does not define overload detection, server-side proof of latest revision, background queue eviction, capacity reservation, deadlines, or recovery control."}],"prior_art_disposition":"ADJACENT_PRIOR_ART","contrastive_claim_remaining":"Under measured overload, a shared monorepo code-intelligence service can safely combine server-verified latest-revision supersession and usefulness deadlines with a protected interactive-capacity reserve, explicit non-retried terminal outcomes, and hysteretic gradual re-entry; relative to admit-all priority scheduling, this combined policy bounds the runnable queue and improves current-revision interactive latency while shedding zero protected tasks.","contrastive_claim_falsifier":"A labelled single-window replay falsifies the claim if shed eligibility cannot be determined before execution with zero protected-task errors, eligible removals fail to bound peak and final runnable-queue size or improve protected-request waiting time versus admit-all priority, discarded work reappears in retry queues, or small threshold perturbations produce unstable repeated shedding and re-admission.","gates":{"adequate_source_search":{"status":"PASS","rationale":"The bounded search covered direct terminology, older and synonymous cancellation/coalescing concepts, documented products and standards, and combinations of queue discard, prioritization, deadlines, and recovery. Exactly four opened sources from four publisher identities were retained, all official, first-party, or standards material.","source_ids":["SRC1","SRC2","SRC3","SRC4"]},"supported_problem":{"status":"PASS","rationale":"Independent sources support both relevant causal components: background indexing can overflow queues and delay user work, and newer edits can make active code analysis stale and cancellable. Evidence for the exact shared-monorepo deployment remains indirect, so problem evidence is only partly supported.","source_ids":["SRC2","SRC4"]},"distinct_testable_claim":{"status":"PASS","rationale":"Known practices cover duplicate rejection, revision cancellation, protocol outcomes, and foreground/background prioritization separately. The remaining combined overload controller is distinguishable through measurable classification accuracy, queue bounds, protected latency, retry behavior, and recovery stability.","source_ids":["SRC1","SRC2","SRC3","SRC4"]},"bounded_next_test":{"status":"PASS","rationale":"A one-service, one-workload-window offline replay can compare identical arrivals under the admit-all priority baseline and shadow shedding, with no production cancellation, while checking zero protected selections, explicit terminal accounting, queue size, latency, released worker time, fairness, and threshold sensitivity.","source_ids":[]},"no_obvious_safety_or_authority_stop":{"status":"PASS","rationale":"The authorized first step is offline shadow evaluation by the service-owning platform team. It excludes integrity and security work, requires independently labelled revision ordering and zero protected-task selections, and makes no live cancellations or externally visible index changes. The LSP caution that older results can remain useful reinforces the required proof-of-supersession safeguard rather than creating a stop.","source_ids":["SRC3"]}},"screen_survival":true,"world_novelty_boundary":"This bounded public-web screen found adjacent prior art for duplicate background-index rejection, revision-triggered cancellation, explicit invalidation responses, and foreground/background queue separation. It did not find the entire proposed overload-triggered combination in the four retained sources, but cannot establish world novelty, patentability, market size, expert acceptance, or realized value; matching implementations may exist in patents, papers, proprietary systems, source code, or documentation outside the searched terminology."}