{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp05_complete_proposal_portfolio20_20260803","cell_id":"layer_decay_and_expiration_management__physics","arm":"COMPLETE_PROPOSAL_PORTFOLIO","candidate_id":"physics_event_pinned_waveform_buffer_lifecycle","proposal_index":5,"version":0,"title":"Precursor-Preserving Eviction for Plasma-Diagnostic Buffers","problem":"A plasma-physics experiment can continuously deposit synchronized waveform chunks from magnetic, optical, particle, and control diagnostics into a fixed-capacity fast buffer. The scientific importance of a chunk may become apparent only after a later disruption, mode transition, or anomaly trigger identifies the preceding interval as a causal precursor. Ordinary first-in-first-out eviction can remove that interval before it is claimed, while indiscriminate pinning can fill the buffer and interrupt acquisition. Confirmed event bundles can also remain indefinitely on fast storage after analysis ends because their dependencies and retention states are unclear.","actors":["plasma physicist responsible for event interpretation","diagnostic physicist responsible for a waveform channel","data-acquisition engineer","experiment session coordinator","timing and synchronization engineer","analysis lead claiming an event window","research data steward"],"observable_state":"For every waveform chunk, the system can inspect diagnostic identity, immutable time interval, sequence number, synchronization quality, calibration reference, creation time, buffer location, access recency, checksum, provisional or confirmed event membership, pre-event and post-event window role, cross-diagnostic bundle links, analysis dependencies, expiry time, storage disposition, and any deletion marker. The failure state is a buffer containing unclassified aging chunks, overlapping pins without owners or release times, silently missing intervals, or confirmed event bundles whose continued fast-tier retention has no explicit purpose.","consequence":"Eviction of an unrecognized precursor interval can leave a later event without the synchronized observations needed to compare the ordering of candidate physical causes. Excessive retention can consume the fast tier and cause admission failures, delayed writes, or loss elsewhere in the stream. If an absent chunk lacks a deletion marker, an analyst may confuse intentional eviction with a diagnostic that measured no signal or failed to produce data.","affected_objective":"Sustain bounded continuous waveform acquisition while preserving complete, synchronized pre-event and post-event windows for explicitly selected plasma phenomena and keeping missing intervals unambiguous.","intervention":"Place the acquisition stream under an event-pinned buffer lifecycle. Divide each diagnostic stream into immutable time chunks with synchronized identities and short default lifetimes. Maintain a rolling look-back horizon in a fixed-capacity fast tier. Under capacity pressure, evict the least-recently needed unpinned chunk, demoting it only when its class permits; an event trigger instead creates a provisional pin over a specified pre-event interval, the triggering interval, a post-event interval, and the matching chunks from required diagnostics. A named reviewer must confirm, resize, or release the provisional bundle before its pin expires. Confirmed bundles are sealed as numbered generations, checked for clock and channel completeness, and moved to the applicable retention path. Before final purge, trace registered analyses and derived products that cite the bundle. Every removed interval receives a compact marker stating whether it expired normally, was evicted under pressure, was corrupt, or was superseded by a preserved bundle. Periodic restore drills verify that sampled archived bundles can be retrieved, parsed, synchronized, and reconnected to their calibration references.","structural_mapping":[{"archetype_element":"Sequential layers accumulate faster than they can remain active","domain_realization":"Continuous diagnostics deposit ordered waveform chunks into a finite fast buffer, so older chunks must leave as new samples arrive."},{"archetype_element":"Older layers acquire different value after deposition","domain_realization":"An ordinary chunk can become a required precursor only when a later plasma event identifies its time interval, while an unconfirmed chunk loses operational value as the look-back horizon advances."},{"archetype_element":"Layer inventory and identity map","domain_realization":"Chunk identities bind diagnostic channel, exact time interval, sequence, synchronization evidence, calibration reference, event membership, and storage state."},{"archetype_element":"Deposition-order and age index","domain_realization":"Sequence numbers and acquisition timestamps establish the generational order used for buffer rotation, expiry, gap detection, and event-window assembly."},{"archetype_element":"Capacity-triggered expiration","domain_realization":"Admission to a full fast tier immediately requires eviction of an eligible unpinned resident; ordinary chunks also expire when their default lifetime or look-back horizon ends."},{"archetype_element":"Preservation exception register","domain_realization":"A provisional or confirmed event pin is an explicit, owned exception that protects a bounded cross-diagnostic interval from ordinary eviction."},{"archetype_element":"Dependency and reconstruction check","domain_realization":"A confirmed bundle cannot be purged while a registered analysis, derived result, calibration investigation, or event comparison cites its raw chunks."},{"archetype_element":"Differentiated disposition paths","domain_realization":"Expired chunks may be evicted with a marker, demoted to a slower tier, incorporated into a confirmed event bundle, compressed, archived, or retained temporarily for integrity review."},{"archetype_element":"Exceptions are revalidated rather than permanent","domain_realization":"Provisional pins expire unless a reviewer confirms them, and confirmed bundles receive class-specific review dates rather than permanent fast-tier residency."},{"archetype_element":"Deletion evidence survives removed data","domain_realization":"Gap markers preserve chunk identity, interval, channel, disposition reason, and bundle successor after waveform bytes are removed."}],"mechanism_mapping":[{"mechanism_slug":"cache_eviction_rule","role":"Controls the fixed-capacity fast tier. When admission finds no free slot, it evicts the least-recently needed eligible chunk while honoring event, calibration, and integrity pins.","counterfactual_removal":"Without an immediate victim-selection rule, buffer pressure produces uncontrolled write failures, manual intervention, or indiscriminate deletion that can remove pinned precursor data."},{"mechanism_slug":"time_to_live_ttl_policy","role":"Stamps ordinary chunks and provisional event pins with lifetimes at creation. Ordinary expiry releases unclaimed chunks; provisional-pin expiry forces confirmation or release rather than creating permanent holds.","counterfactual_removal":"Without preassigned lifetimes, ordinary chunks and speculative pins remain resident until a cleanup sweep, allowing forgotten material to consume the acquisition budget."},{"mechanism_slug":"log_rotation_and_cleanup_job","role":"Seals confirmed event bundles and ordinary stream segments into numbered time generations, compresses or archives eligible generations, and purges the tail according to class rules while preserving sequence continuity.","counterfactual_removal":"Without ordered rotation, chunks leave the buffer as unrelated files, making temporal gaps, cross-channel alignment, and bounded retention harder to verify."},{"mechanism_slug":"retention_schedule","role":"Assigns distinct retention and review paths to untriggered background, provisional events, confirmed physics events, calibration intervals, commissioning data, and corrupt or incomplete bundles.","counterfactual_removal":"Without class-specific rules, every trigger either receives the same retention as routine background or becomes an indefinite exception, overwhelming the distinction created by event pins."},{"mechanism_slug":"stale_layer_detection_dashboard","role":"Shows buffer occupancy, chunk age, pin owner and deadline, synchronization gaps, incomplete event bundles, fast-tier residents that should have rotated, and preserved bundles with expired review dates.","counterfactual_removal":"Without a visible inventory, orphaned pins and incomplete cross-diagnostic bundles can silently consume capacity or appear scientifically complete."},{"mechanism_slug":"dependency_safe_delete_check","role":"Gates final purge of confirmed bundles by tracing registered analyses, derived event features, comparison sets, and calibration investigations that reference their raw chunks.","counterfactual_removal":"Without the dependency gate, a bundle can be removed after its initial review even though a downstream analysis still depends on its raw timing or waveform detail."},{"mechanism_slug":"archive_restore_test","role":"Samples archived event bundles across age, diagnostic composition, and format, then exercises retrieval, parsing, synchronization, checksum validation, and reconnection to the recorded calibration references.","counterfactual_removal":"Without end-to-end restoration, successful archival cannot establish that a preserved event window remains usable as a synchronized physics record."},{"mechanism_slug":"tombstone_or_deletion_marker","role":"Leaves an interval marker after eviction or purge, recording channel, time range, reason, decision path, and any preserved bundle that supersedes the deleted chunk.","counterfactual_removal":"Without markers, downstream users cannot distinguish intentional expiration from a zero signal, acquisition failure, corruption, or missing transfer."}],"causal_chain":["Continuous diagnostic sampling deposits time-ordered waveform chunks into a fixed-capacity buffer.","Default lifetimes and sequence identities make every chunk's ordinary residency and expected expiration inspectable.","A later event trigger reaches backward into the still-resident horizon and provisionally pins the precursor, event, and post-event chunks across required diagnostics.","The eviction rule continues admitting new data by removing only eligible unpinned chunks rather than freezing the entire buffer.","Reviewer confirmation releases false or low-value pins and seals selected cross-diagnostic intervals as complete event generations.","Completeness and synchronization checks expose missing channels or timing uncertainty before the bundle is treated as an analyzable causal window.","Rotation moves confirmed bundles off the fast tier while class rules govern compression, archive review, and eventual purge.","Dependency checks preserve bundles still supporting registered analyses, and restore drills test the far end of archival recoverability.","Gap markers make every removed interval explicit, while expiring pins and rotating generations keep the acquisition stack bounded."],"baseline":"The baseline uses a circular first-in-first-out waveform buffer plus hardware or software triggers that copy selected data to storage. Trigger definitions, buffer depth, manual file retention, and later archival are managed separately. A delayed trigger may find its precursor overwritten, copied event files may omit synchronized channels, and trigger outputs can remain on fast storage without owned expiry or explicit gap markers.","nearest_rivals":["A deeper first-in-first-out buffer: extends the look-back interval but does not distinguish event-bearing chunks from ordinary residents or prevent eventual precursor overwrite.","Trigger-only acquisition without a rolling buffer: bounds storage but cannot preserve observations preceding a trigger that was recognized later.","Write every waveform directly to long-term storage: avoids selective eviction but transfers the unbounded stream into downstream storage, search, and maintenance obligations.","Fixed periodic downsampling: reduces volume but can discard transient timing or waveform structure before an event is recognized.","An anomaly detector that immediately archives selected channels: automates selection but can omit required cross-diagnostic context and does not govern false-trigger release or bundle expiry.","Manual post-run file cleanup: occurs after acquisition pressure has already forced losses and cannot recover overwritten precursor chunks."],"remaining_contrastive_claim":"The proposal's specific claim is that delayed scientific recognition can be reconciled with a hard acquisition ceiling by combining a rolling look-back buffer, bounded cross-diagnostic event pins, immediate unpinned eviction, generational sealing, and explicit gap markers. Its core intervention acts during data admission and event formation, rather than curating an already completed scientific archive or expiring the authority of executable content.","authority_safety":{"decision_authority":"The data-acquisition lead controls buffer capacity and eviction implementation; diagnostic owners define required channels and integrity constraints; the experiment session coordinator may initiate or confirm provisional event pins; the analysis lead registers downstream dependencies; the data steward authorizes long-term retention and purge classes. No anomaly score alone may declare an event scientifically valid.","authorized_first_step":"Replay a copied, non-authoritative waveform trace through an isolated buffer simulator using recorded or deliberately injected trigger delays. Compare the proposed event-pin lifecycle with the existing first-in-first-out, fixed-lifetime, and trigger-only rules. Do not connect the simulator to live acquisition, alter source files, or change production buffer settings.","excluded_actions":["deploying eviction or pinning changes to a live diagnostic during the first evidence step","purging or rewriting source waveform files","treating automatic trigger output as confirmed scientific classification","pinning an unbounded time range or unrestricted set of channels","altering timestamps or synchronization metadata to make a bundle appear complete","releasing a registered calibration or analysis dependency without its owner","using an offline result as production authorization without the normal data-acquisition review"],"halt_rollback":"Halt if replay cannot preserve original sequence and timestamps, the simulator reads from or writes to production paths, a known event window is classified as safely evictable, or pinning causes unbounded simulated occupancy. Preserve the copied input unchanged, discard only isolated simulation outputs, and withdraw all simulated retention labels from operational use."},"negative_tests":{"strongest_counterevidence":"All scientifically relevant triggers are available before their earliest required precursor chunk approaches eviction, the baseline capture already preserves complete synchronized channel sets within the fixed capacity, and retained trigger bundles are explicitly reviewed and purged. Under those conditions, event-pinned lifecycle control adds no missing causal capability.","problem_falsifier":"The problem is falsified for the pilot trace if every adjudicated event has its required precursor and post-event chunks under the baseline, no buffer admission failure or ambiguous gap occurs, and existing event bundles already have explicit owners, dependencies, and expiry states.","intervention_falsifier":"The intervention is falsified if provisional pins routinely exhaust the simulated fast tier, confirmed bundles remain incomplete or mis-synchronized, known event windows are evicted, ordinary acquisition continuity is not preserved, or the lifecycle provides no decision-relevant improvement over a simple increase in buffer depth or first-in-first-out retention.","risks":["A burst of provisional triggers can pin more data than the fast tier can hold.","Clock skew can associate the wrong chunks across diagnostics.","A delayed or incorrect trigger can preserve irrelevant data while allowing a useful interval to expire.","Priority rules can systematically favor some diagnostics or event classes without scientific justification.","Compression or format conversion can alter usable waveform precision or detach calibration references.","Gap markers can be mistaken for proof that all surrounding data are complete.","A registered analysis dependency may persist after its owner stops reviewing it.","Successful restoration of sampled bundles does not establish recoverability of every archived generation.","Event metadata and markers can accumulate if their own retention rules are not enforced."]},"next_evidence_step":"Select a bounded copied trace containing ordinary operation, calibration intervals, complete events, delayed triggers, overlapping provisional triggers, synchronization faults, and buffer-pressure episodes. Replay identical inputs through first-in-first-out, fixed-lifetime, trigger-only, and event-pinned policies at the same simulated capacity. Have an independent plasma-and-diagnostics panel define the required time and channel window for each adjudicated event before seeing retention outcomes. Measure whether each policy preserves those windows, sustains admission, releases false pins, produces explicit gaps, and forms restorable bundles. Restore a sample of sealed bundles in isolation. Make no production or source-data changes.","prior_art_status":"UNSEARCHED","diversity_from_prior_proposals":"Proposal 1 manages physical debris films on operating optical shields through condition review, carrier replacement, intact quarantine, and cleaning or disposal. This replacement manages ephemeral waveform chunks under real-time buffer pressure; it contains no physical surface renewal or deposited-material preservation. Proposal 2 manages completed Markov-chain configurations through statistical qualification, restart-aware storage tiering, and long-term archival disposition. This proposal instead acts at acquisition admission time, where an immediate cache victim must be selected before a later trigger can claim a bounded precursor window; it neither determines ensemble membership nor preserves restart states. Proposal 3 expires the default resolver authority of calibration payloads. This proposal's lifetimes govern byte residency and provisional buffer pins, not scientific applicability or executable authority. Proposal 4 manages finite irradiated specimens through physical custody, aliquot reserves, destructive consumption, and disposal. This proposal manages reproducible digital time chunks with no material handling, mass balance, radiological state, or specimen depletion. It is independently adoptable by a plasma-diagnostic acquisition team without adopting any other current proposal.","revision_record":{"parent_version":null,"progress_targets_addressed":["Replaced the prior index-5 candidate after the auditor found its physical-deposit renewal intervention insufficiently distinct from proposal 1.","Created an independently adoptable hard-capacity acquisition opportunity materially different from proposals 1 through 4.","Preserved proposal index 5 and version 0.","Specified complete actors, observable state, consequence, intervention, mechanism mapping, causal chain, rivals, authority, safeguards, falsifiers, and bounded evidence."],"conceptual_changes":["Replaced cryopanel sorbate accumulation with sequential waveform chunks in a fixed-capacity diagnostic buffer.","Changed the affected objective from vacuum-surface regeneration to continuous acquisition with precursor-window preservation.","Made delayed event recognition, bounded pins, and capacity-triggered eviction the central causal structure."],"operational_changes":["Replaced physical isolation and warming with chunk admission, provisional cross-diagnostic pinning, immediate unpinned eviction, generational sealing, and explicit gap markers.","Restricted initial evidence to deterministic replay of copied traces in an isolated buffer simulator."],"evidence_changes":["Added same-capacity comparisons against first-in-first-out, fixed-lifetime, trigger-only, and deeper-buffer alternatives.","Specified independent event-window adjudication and isolated restoration without assuming an effect size."],"claim_changes":["Withdrew the prior claim about campaign-indexed sorbate regeneration.","Limited the replacement claim to precursor-preserving lifecycle control under a hard acquisition ceiling.","Made no claim of novelty, prevalence, demand, or effect size."]}}