Independent source-extraction pipelines recover the canonical detections (eight of nine from April 1950; three of three from July 1952) and roughly two-thirds of the published 5,399-candidate catalog, without reference to the original team's software or candidate lists.

Hayes's benchmarked recovery rates settle software-dependence; the same work shows the candidate population's size is a function of analysis choices, which disciplines every ensemble claim.

Independent source-extraction pipelines successfully recover the canonical Palomar marks and the majority of the published candidate catalog from the raw digitizations without referencing the proponents' code. The strongest opposing consideration is that this only proves reproducibility of detection within the shared digitized descendants, which would equally recover a persistent population of copy defects.

Hayes’s independent detection pipeline, developed without access to the VASCO software or candidate list, recovers eight of nine 1950 sources, all three 1952 sources, and about two-thirds of the public 5,399-candidate catalog, with documented matching statistics and injection tests (TRN-E008). This clearly shows that the canonical detections and a substantial fraction of the catalog are not artifacts of a particular codebase. No contrary evidence suggests that other independent pipelines fail to recover these when applied to the same data. The strongest opposing consideration is that both pipelines operate on the same digitized copy lineage, so while software independence is established, physical independence is not.

Hayes's from-scratch pipeline, with synthetic-injection controls, recovers eight of nine 1950 sources, three of three 1952 sources, and 3,450 of Solano's 5,399 published candidates, which kills a one-team software-bug explanation. The same work shows a 500-fold swing between a loose full-footprint sweep and the vetted public list, so population size is not a stable quantity. The strongest opposing consideration is that reproducible detection in shared digitized copies is exactly what a defect population would also produce.

Hayes's from-scratch pipeline, benchmarked with synthetic injections and control fields (~0.2 false positives per control field), recovers 8 of 9 April 1950 sources, 3 of 3 July 1952 sources, and 63.9% of the published 5,399-candidate catalog at sub-arcsecond median separation. That cleanly kills the "one team's software bug" hypothesis. The strongest opposing consideration is the flip side documented in the same work: his full-footprint sweep yields 2.85 million candidates — a roughly 500-fold spread showing the population is a function of extraction and vetting choices — and independent extraction still reads the same digitized descendants of the same photographic copies, so reproducible detection is exactly what an inherited defect population would also produce.

Hayes documents concrete recovery rates (8/9, 3/3, 3,450 of 5,399 matched at 0.94 arcsec median separation) with control fields and synthetic injections, which cleanly kills the "one team's software bug" explanation for the canonical detections. Confidence is capped at moderate because the claim's plural framing ("pipelines") rests on exactly one single-author preprint run with acknowledged preliminary parameters. The strongest opposing consideration is what the same work reveals rather than what it confirms: a 2.85-million-candidate sweep against a vetted 5,399 shows the population's composition is almost entirely a function of extraction and vetting choices, so reproducible detection here demonstrates data fidelity, not physical origin.

TRN-E008's benchmarked recovery rates settled this; nothing in the new material touches it. The 500-fold sensitivity of catalog size to extraction choices continues to discipline every ensemble claim above it.

TRN-E008 settled software-independence with benchmarked recovery rates while exposing how strongly the candidate population depends on extraction and vetting choices. Unchanged.

TRN-E008 settles software-independence with benchmarked false-positive rates; the same work shows the candidate population's size swings 500-fold with extraction choices, which disciplines every ensemble claim built on it.

TRN-E008 settles software-dependence with benchmarked recovery rates, while showing the candidate population's size is a function of extraction and vetting choices — which disciplines every ensemble claim above it.

Hayes's 2026 independent, from-scratch detection pipeline successfully recovers nearly all of the canonical marks (eight of nine from 1950, three of three from 1952) and roughly 64% of the public candidate catalog. This robustly demonstrates that the detections are reproducible features of the digitized data rather than artifacts of the original team's private software code. The strongest opposing consideration is that reproducible recovery from the same shared digitizations proves only software reliability; a population of shared copy-defects or emulsion flaws would be identically recovered by any competent pipeline.

An independently constructed extraction pipeline successfully recovers nearly all the canonical events and approximately 64% of the published candidate catalog. This definitively proves the detections are not software bugs unique to the original authors. The strongest opposing consideration is that the pipeline reads the same digitized descendants of the photographic copies, meaning it reproduces the extraction but cannot differentiate between physical sky events and shared copy defects.

Hayes’s independent source-extraction pipeline, developed from scratch without using the VASCO software or candidate lists, recovers 8 of 9 1950 sources, all 3 of 1952, and about two-thirds of the 5,399-object public catalog, validated with synthetic injections and control fields (TRN-E008). This shows that the canonical detections and a large fraction of the catalog are reproducible across algorithmic choices within the same digitized image set. The strongest opposing consideration is that this independence is limited to software; both pipelines operate on scans descended from the same photographic copies, so physical origin remains unaddressed.

Hayes’s from-scratch pipeline, with no access to the VASCO team’s codes or candidate list, recovers 8/9 of the 1950 sources, 3/3 of the 1952 triple, and roughly two-thirds of the published 5,399-candidate catalog with sub-arcsecond agreement (TRN-E008). Synthetic injections and control fields show low false-positive rates in small regions, indicating the canonical detections are not software artifacts unique to the original pipeline. The strongest opposing consideration is that both pipelines operate on the same digitized plate products, so agreement does not address whether the underlying marks originate on the original negative or in later copying or scanning stages.

Hayes's independently written pipeline, with synthetic-injection controls, recovers eight of nine of the 1950 event, three of three of the 1952 event, and 3,450 of Solano's 5,399 published candidates (TRN-E008). That kills a one-team software-bug account of the detections. The strongest opposing consideration is that reproducible extraction from shared digitized copies is exactly what a defect population would also produce, and Hayes's loose full-sky sweep exploding to 2.85 million candidates shows how little headline catalog size means.

Hayes, building a pipeline from scratch with synthetic-injection controls, recovers eight of nine April 1950 sources, three of three July 1952 sources, and 3,450 of Solano's 5,399 published candidates at a median separation of 0.94 arcsec. A software-specific artifact of the original team's code is ruled out. Strongest opposing consideration: independent extraction from shared digitized copies is exactly what a defect population would also produce, and Hayes's full sweep yielded 2.85 million candidates against 5,399, so the population's headline size is a function of cuts rather than a sky census.

Hayes's from-scratch pipeline — sigma-clipped detection, PSF filtering, cross-epoch matching, synthetic-injection benchmarks — recovers 8/9 of the April 1950 sources, 3/3 of the July 1952 triple, and 63.9% of Solano's 5,399 published candidates at sub-arcsecond median separation (E008), cleanly retiring "one team's software bug" as an account of the detections. The claim as stated is about recovery, and its documentation is strong for a single-author preprint. The strongest opposing consideration is that independent extraction reads the same digitized descendants of the same copies, so it cannot bear on physical origin — and the same sweep produced 2.85 million raw candidates, a 500-fold spread showing how vetting-dependent the population is. That limits the claim's diagnosticity, not its truth. Well supported, high confidence.

Hayes 2026 documents the recovery rates — eight of nine from 1950, three of three from 1952, 63.9% of the published 5,399-candidate catalog — with benchmark control fields and synthetic-injection tests, which cleanly kills the "one team's software bug" hypothesis. The claim as stated is about reproducible detection in shared imagery, and that is demonstrated. The strongest opposing consideration is scope: the independent pipeline reads the same digitized descendants of the same photographic copies, so it cannot speak to physical origin, and the 500-fold spread between his 2.85-million-candidate sweep and the vetted 5,399 shows how completely the "population" depends on extraction choices. That undermines ensemble claims built on the catalog, not the recovery result itself.

Hayes documents 8/9, 3/3, and 3,450 of 5,399 (63.9%) at median 0.94 arcsec separation, with control-field false-positive rates and synthetic injection tests (TRN-E008) — this cleanly retires the "one team's software bug" explanation. Confidence is moderate because it is a single-author preprint whose full-sky sweep admittedly ran with loose preliminary parameters, and no one has re-run it. The strongest opposing consideration is that the claim is far less probative than it first appears: independent extraction from the same digitized descendants of the same photographic copies is exactly what a defect population would also reproduce, and the 500-fold spread between the 2.85-million sweep and the vetted 5,399 shows how much the "population" is a product of vetting choices.

Hayes documents 8/9, 3/3, and 63.9% catalog recovery at median 0.94 arcsec separation, with ~0.2 false positives per control field and synthetic-injection benchmarks (TRN-E008), which cleanly retires "one team's software bug." Two qualifiers hold confidence at moderate: this is a single unreviewed preprint by a single author, so "pipelines" plural overstates the record, and the same work shows the population is extraction-dependent to a startling degree (2.85 million candidates against 5,399). The strongest opposing consideration for the case, though not for the claim, is that reproducible detection in shared digitized imagery is exactly what a defect population would also yield — this claim's local truth carries almost no upward diagnostic load.

TRN-E008 demonstrates software independence with benchmarked recovery rates, while TRN-E016 shows how strongly catalog size depends on extraction and vetting choices. No new material disturbs this.

TRN-E008 settles software-independence with benchmarked recovery rates while showing the catalog's size is a function of extraction and vetting choices. Nothing in the new batch touches this rung.

TRN-E008 demonstrates software independence with benchmarked recovery rates while exposing how strongly catalog size depends on extraction and vetting choices (TRN-E016). Nothing in the current material disturbs this.

TRN-E008 demonstrates software independence with benchmarked recovery rates while exposing how strongly catalog size depends on extraction and vetting choices (TRN-E016). No new material disturbs this.

An independent extraction pipeline successfully rebuilt a catalog from scratch that recovered the canonical events and most of the public candidate dataset, ruling out simple software bugs. The strongest opposing consideration is that reproducing detections from the same underlying digitized copies does not prove physical origin, and the vast size of the raw background highlights sensitivity to vetting choices.

Hayes’s independent pipeline, using different detection, PSF filtering, and cross-matching choices, recovers most canonical events and approximately two-thirds of the published 5,399-candidate catalog (TRN‑E008), with injection tests to estimate false positives. This robustly shows that the existence of candidate marks is not an artifact of one team’s code. The strongest opposing consideration is that all pipelines analyze the same underlying digitized copies, so reproducible detection does not resolve whether the marks are real flashes or plate defects.

Hayes 2026, built from scratch with sigma-clipped detection, PSF filtering, and cross-epoch rejection, recovers eight of nine April 1950 sources and three of three July 1952 sources, and matches 3,450 of Solano’s 5,399 published candidates (63.9%) at a median 0.94 arcsec. That kills the “one team’s private code” objection. The strongest opposing consideration is that reproducible detection in shared source imagery is exactly what a defect population would also produce, and Hayes’s full-footprint sweep generated 2.85 million candidates against Solano’s 5,399 — a 500-fold swing that shows how little the population’s headline size means and how completely every later ensemble statistic inherits its extraction cuts.

Hayes's from-scratch pipeline recovers 8 of 9 April 1950 sources, 3 of 3 July 1952 sources, and 63.9% of the published 5,399-candidate catalog, with benchmarked false-positive rates — the detections are not one team's software artifact. The claim as stated is about recovery, and recovery is demonstrated with controls and synthetic injection. The strongest opposing consideration is that this is independent extraction from shared material: the pipeline reads the same digitized descendants of the same copies, so a defect population would reproduce exactly this result — and the 500-fold spread between Hayes's 2.85 million raw candidates and the vetted 5,399 shows how fluid the "population" is. That limits diagnosticity, not the claim's truth.

Hayes documents recovery rates with benchmark controls, synthetic injections, and a quantified false-positive rate (~0.2 per 10-arcmin control field), and the 63.9% match to Solano's published catalog at 0.94 arcsec median separation is specific and checkable. This cleanly kills the "one team's software bug" explanation. Confidence is moderate because the claim says "pipelines" plural while the record contains one unreviewed single-author preprint, and because the full-sky sweep ran with deliberately loose parameters the author himself flags. The strongest opposing consideration is diagnostic rather than evidential: reproducible detection of the same marks in the same shared digitized imagery is precisely what a defect population would also produce, and the 500-fold spread between 2.85 million and 5,399 candidates shows how little the population's size means.

TRN-E008 demonstrates software independence with benchmarked recovery rates, while TRN-E016 shows how strongly catalog size depends on extraction and vetting choices. No new material disturbs this.

The Hayes independent detection pipeline successfully recovers the canonical events and approximately 64% of the public catalog (TRN-E008), proving the detections are not an artifact of the proponents' proprietary software. The pipeline demonstrates robust recovery against synthetic injections. The strongest opposing consideration is that this independent full-sky sweep also generated 2.85 million candidates, highlighting that the "transient population" size and composition are overwhelmingly dependent on subjective extraction and vetting choices.

Hayes’s independent pipeline, built without access to VASCO’s software or candidate lists, recovers 8/9 of the 1950 sources, 3/3 of the 1952 triple, and about two-thirds of the published 5,399-candidate catalog in the same digitized data (TRN-E008). This convincingly shows that the canonical detections and a large fraction of the catalog are not artifacts of a single team’s code or parameter choices. The strongest opposing consideration is that both pipelines still operate on the same digitized descendants of the same plates, so they cannot distinguish real sky events from plate or copy defects, but that does not weaken the methodological replication claim itself.

Hayes's from-scratch pipeline, with synthetic-injection controls, recovers eight of the nine 1950 sources, all three 1952 sources, and 3,450 of Solano's 5,399 published candidates at a median 0.94 arcsec. A private-software account of the detections is ruled out. Strongest opposing consideration: reproducible detection in shared digitized copies is exactly what a defect population would also produce, and the 500-fold gap between his 2.85-million-candidate sweep and the vetted 5,399 shows that headline population size is a vetting choice.

Hayes's from-scratch pipeline, benchmarked with synthetic injections and control fields, recovers eight of nine 1950 sources, three of three 1952 sources, and roughly 64% of the public catalog. The strongest opposing consideration is that independent software still reads the same digitized descendants of the same copies — so this kills "one team's software bug" cleanly but cannot touch physical origin — and the 500-fold spread between the full sweep and the vetted catalog shows how unstable the population definition is. As a claim about recoverable detections in shared imagery, it is solid.

TRN-E008 documents recovery of 8/9 and 3/3 canonical sources plus 63.9% of the published 5,399-candidate catalog at 0.94″ median separation, with synthetic-injection benchmarking and ~0.2 false positives per control field — this cleanly kills "it is one team's software bug." Confidence is moderate because this is a single unreviewed preprint by one researcher and the pluralized "pipelines" of the claim is really one pipeline. The strongest opposing consideration is the flip side reported in the same work: a full-footprint sweep produced 2.85 million candidates against 5,399, so reproducible detection of marks in shared imagery is exactly what a defect population would also produce, and the "population's" size is a function of vetting choices rather than of the sky.

TRN-E008 demonstrates software independence with benchmarked false-positive rates; TRN-E016 shows how extraction choices govern catalog size. Note this is independent extraction, not independent plate material.

TRN-E008 demonstrates software independence with benchmarked false-positive rates; TRN-E016 shows how extraction choices drive catalog size. Unchanged.

TRN-E008 demonstrates software independence with benchmarked false-positive rates; TRN-E016 shows the catalog's size is extraction-choice dependent. Nothing in the current material disturbs this.

TRN-E008 demonstrates software independence with benchmarked false-positive rates; TRN-E016 shows how extraction choices govern catalog size. No new material disturbs this.

TRN-E008 demonstrates software independence with benchmarked false-positive rates; TRN-E016 shows the population's size is extraction-choice dependent. Unchanged.

TRN-E008 demonstrates software independence with benchmarked recovery rates, while TRN-E016 shows how strongly catalog size depends on extraction choices. Nothing in the current material disturbs this.

TRN-E008 demonstrates software independence with benchmarked false-positive rates; TRN-E016 shows how extraction choices govern catalog size. Nothing new disturbs this.

TRN-E008 demonstrates software independence with benchmarked recovery; TRN-E016 shows catalog size is extraction-choice dependent. Nothing new disturbs this.

The Hayes pipeline definitively demonstrates that independent, from-scratch software can recover the canonical events and a majority of the published catalog without relying on the VASCO team's proprietary code. However, because this pipeline extracts from the identical digitized copy descendents, it cannot determine whether it is reliably extracting celestial flashes or reliably extracting shared plate defects.

Hayes’s fully independent pipeline, with its own detection and cross-matching logic, recovers almost all canonical events and about two-thirds of the published 5,399-candidate catalog, demonstrating that these marks are not artifacts of one team’s software. Synthetic injection tests support the pipeline’s basic performance. Replication across software on shared imagery is exactly what this claim asserts. The strongest opposing consideration is that this independence is methodological, not material: both pipelines still analyze the same digitized copies and thus cannot discriminate between real transients and plate or copy defects.

Hayes's from-scratch pipeline recovers eight of nine 1950 sources and three of three 1952 sources, matches 63.9 percent of Solano's 5,399 published candidates, and documents false-positive and injection-recovery controls. That kills the hypothesis that the detections are an artifact of one team's private code. The strongest opposing consideration is that independent software on shared digitized copies is exactly what a defect population would also produce, and the 500-fold spread between Hayes's sweep and the vetted public catalog shows how little headline population size means.

Hayes's from-scratch pipeline, benchmarked with controls and synthetic injections, recovers the canonical events and 63.9% of the published vetted catalog (TRN-E008), cleanly killing the "one team's software bug" hypothesis. The claim is about recovery, and recovery holds. The strongest opposing consideration is that independent code reads the same digitized descendants of the same copies — so this cannot speak to physical origin — and the same work's 2.85-million-candidate sweep shows the population's size is an artifact of extraction choices; both points qualify the population's meaning, not the recovery fact.

TRN-E008 documents specific, checkable numbers — 8/9, 3/3, 3,450 of 5,399 matched at 0.94 arcsec median separation, ~0.2 false positives per control field, synthetic-injection benchmarks — from a pipeline built without reference to the original team's software, which cleanly kills the "one team's code bug" hypothesis. Confidence is moderate because this rests on a single unreviewed preprint by one author whose full-sky sweep admittedly ran with preliminary, looser parameters. The strongest opposing consideration is directional rather than factual: reproducible detection in shared source imagery is exactly what a defect population would also yield, and the 500-fold gap between Hayes's 2.85 million candidates and Solano's 5,399 shows the "population" is defined by vetting choices, not by the sky.

TRN-E008 demonstrates software independence with benchmarked recovery rates; TRN-E016 shows the catalog's size is extraction-choice dependent. No new material disturbs this.

TRN-E008 demonstrates software independence with benchmarked recovery rates; TRN-E016 documents how catalog size depends on extraction choices. Nothing new disturbs this.

Hayes 2026 documents the recovery rates with benchmark controls and synthetic-injection tests. The same work shows the flip side: his full-sky sweep produced 2.85 million candidates against Solano's 5,399 — the population's size and composition swing enormously with extraction and vetting choices.

An independent pipeline utilizing sigma-clipped detection, point-spread function filtering, and cross-epoch rejection successfully recovered eight of nine April 1950 sources, three of three July 1952 sources, and approximately 64% of the published 5,399-candidate catalog. This demonstrates that recovery of the primary candidate catalog does not rely on proprietary or flawed software algorithms from the discovery team. The strongest opposing consideration is that the same independent sweep produced nearly three million raw candidates under relaxed parameters, emphasizing the catalog's extreme sensitivity to threshold vetting choices.

TRN-E008 (strong) recovers the canonical events with injection-recovery benchmarking and matches 63.9% of the published catalog, demonstrating the detections do not depend on the VASCO team's software. Strongest opposing consideration: the replication reads the same digitized DSS descendants of the same photographic copies, so it is independent extraction but not independent material, and the full-sky sweep ran with loose preliminary parameters its author flags.

TRN-E008 reports a from-scratch pipeline recovering eight of nine April sources, all three July sources, and 3,450 of 5,399 catalog candidates, or 63.9 percent. This directly supports the numerical core of the claim and shows that the detections do not depend on the original software. The strongest opposing consideration is that only one clearly independent extraction pipeline is documented and its loose full-sky run produced 2.85 million candidates, exposing strong pipeline dependence in the broader population.

Hayes’ from-scratch pipeline recovers 8/9 of the 1950 sources and 3/3 of the 1952 sources and matches 63.9 % of the published 5,399- candidate list at ~1 arcsec, with low control-field false-positive density. Independence is of software and candidate lists, not of plate material. Strongest opposing consideration: the same pipeline’s loose full-footprint sweep yields millions of candidates, showing how strongly catalog size tracks vetting choices.

TRN-E008 is the case's cleanest methodological result: a pipeline built without reference to the VASCO software or candidate lists recovers 8/9, 3/3, and 63.9% of the published 5,399, at median 0.94 arcsec, with ~0.2 false positives per control field. The specific numbers in the claim match the source. The qualification, which the claim's wording already accommodates, is that this is independent extraction from the same digitized descendants rather than independent material, and that the full-sky sweep used deliberately loose parameters — yielding 2.85 million candidates and thereby showing how strongly "the population" depends on vetting choices.

TRN-E008 (strong): a from-scratch pipeline recovers 8/9 and 3/3 canonical sources and 63.9% of the 5,399 catalog with ~0.2 false positives per control field. TRN-E016 gives the catalog construction context. The same record shows a 2.85-million-candidate full-sky sweep against a vetted 5,399, demonstrating how completely the 'population' depends on extraction and vetting choices.

TRN-E008 (strong): from-scratch pipeline recovers the canonical events and 63.9% of the 5,399 catalog. TRN-E016 supplies the catalog-construction context.

Hayes developed an independent source-detection pipeline from scratch that successfully recovered 8 of the 9 April 1950 sources, all 3 July 1952 sources, and 63.9% of Solano's published 5,399-object catalog. This proves that candidate recovery does not rely on private code or idiosyncratic parameters of the VASCO group. Strongest opposing consideration: Hayes's full-sky sweep yielded 2.85 million raw candidates, emphasizing that catalog membership is extremely sensitive to thresholding and vetting cuts.

TRN-E008 delivers exactly what the claim describes: 8/9 and 3/3 canonical recovery with a benchmarked false-positive rate, and 63.9% catalog recovery without reference to the original software. Strongest opposing consideration: this is independent extraction of the same digitized lineage, not independent material, and the author's own 2.85-million-candidate preliminary sweep shows the "population" is overwhelmingly a function of vetting choices — the 5,399's meaning is not secured by recovery alone.

TRN-E008 directly reports a from-scratch pipeline recovering eight of the nine April candidates, all three July candidates, and 3,450 of 5,399 published catalog entries. This closely matches every numerical component of the claim without relying on the original team's software. The strongest opposing consideration is that the pipeline used the same digitized photographic lineage and generated 2.85 million loose candidates, so recovery does not validate the physical reality of the detections.

An independent from-scratch pipeline recovers 8/9 of the 1950 sources and 3/3 of the 1952 sources and matches 63.9 % of the published 5399-candidate list at sub-arcsecond median separation (E008). Strongest opposing consideration is that the pipeline still reads the same digitized copy lineage, so it does not constitute independent material.

Hayes's pipeline, built from scratch, recovers eight of nine and three of three canonical sources at ~0.2 false positives per control field. It matches 63.9% of the 5,399 published candidates (E008), without using VASCO software or lists. Strongest opposing consideration: the claim says "pipelines", but this is one single-author preprint. Its catalog match used preliminary parameters (2.85 million candidates), and recovery stays within the same digitized copy lineage.

TRN-E008 (strong) describes a from-scratch pipeline. It recovers 8 of 9 and 3 of 3 canonical sources and 63.9% of the 5,399 catalog, with low false positives in control fields. TRN-E016 is context on how that catalog was built. The software is independent; the digitized material is the same.