PREreview del The U.S. Interconnection Queue System: Cascading Vulnerability Analysis and a Resilience Engineering Framework

Review of “The U.S. Interconnection Queue System: Cascading Vulnerability Analysis and a Resilience Engineering Framework” (arXiv:2609.10455v1)

Summary. The paper argues that interconnection queues are vulnerable to cascades of withdrawals driven by restudies and the reallocation of network upgrade costs. It tests this with dispersion, burst and permutation analyses of LBNL Queued Up 2026 Edition data for seven ISO/RTOs (2010–2025), and with a stylised Erdős–Rényi contagion model including circuit-breaker caps. The paper is careful in many places: the sensitivity checks in Table IV, the entry-year-stratified null, and the treatment of the k = 20 result as a boundary condition are all welcome. My main comments concern whether some interpretive claims go further than these results allow.

1. The 2025 wave and Experiment 7. Section IV-D states that the roughly 25% of active projects withdrawn in 2025 “corresponds to” Experiment 7, which yields 38.5% total failure at k = 10. However, Experiment 7 takes the 2025 annual wave (24.9%) as its input shock. The observed wave therefore cannot validate the model’s output. If anything, the model’s 38.5% exceeds the roughly 25% observed. I would suggest rewording this passage, or comparing the model against an observation not used to set its inputs.

2. The “out-of-sample test” does not separate the two explanations. Section IV-D presents 2025 as an out-of-sample test of the prediction that withdrawals cluster around reform deadlines. But the paper’s own first interpretation, screening of speculative projects by Order 2023 readiness requirements, predicts the same clustering without any contagion. A test that both explanations pass cannot favour one. The section would be stronger if it stated what pattern would distinguish them, for example whether withdrawals follow restudy results that raised a project’s costs.

3. A technology-specific common shock in 2025. The technology-label permutation finds co-withdrawal concentrated within technology categories in all seven regions. In 2025, the reconciliation act (Public Law 119-21) accelerated the termination of the clean electricity tax credits for wind and solar. That is a shock specific to technology categories, falling within the study period, and the paper does not mention it. I would suggest repeating the technology test with 2025 excluded, or comparing wind and solar with storage and gas, to see whether the concentration survives.

4. The heavy-tail interpretation. Section IV-C states that the heavy-tailed distribution “suggests that relatively small perturbations may trigger disproportionately large withdrawals”. Section IV-B, however, notes that the pooled CCDF “may then partly arise from heterogeneity in the underlying distributions across individual ISO/RTOs”, and the largest bursts are linked to cluster deadlines and bulk-assigned dates, which are large administrative events rather than small perturbations. The Section IV-C sentence should be qualified to match. Likewise, the closing sentence of Section IV-B (“These findings are consistent with a contagion effect”) would be more accurate if it added that they are equally consistent with common shocks, as the preceding sentences acknowledge.

5. Missing withdrawal dates. About 2,222 of 16,206 withdrawals (13.7%) are undated. The paper treats the ERCOT and NYISO indices as lower bounds, but this holds only if the missing dates would have added clustering. If dates were recorded selectively, as the paper suggests they may have been, the bias could run either way. SPP’s coverage (72.8%) is also well below the other regions but is not flagged. Comparing dated and undated withdrawals by entry year, technology and size would show whether the missingness is likely to matter.

6. Scope. The title refers to the U.S. interconnection queue system, and the introduction notes that the dataset covers more than 50 transmission operators, but the analysis covers only the seven ISO/RTOs (25,770 of 38,201 records). Utilities outside organised markets are excluded without comment. I would suggest stating this scope in the abstract and explaining the exclusion. Withdrawal-date coverage outside the ISOs is considerably lower, which may be the reason.

Minor points.

· The “Excluding Imputed Dates” column of Table IV refers to a “date-imputation procedure” that is not described anywhere in the paper and note (b) describes the same column as excluding bulk-assigned dates.

· Please define the procedure and reconcile the two descriptions.

· “Cluster-dated” in the abstract is not defined; the text uses “cluster-deadline”.

· Section V-D notes that the transition is governed by kα and scales with 1/α. Given α = 0.08, the self-sustaining regime begins at kα = 1, that is k = 12.5, by construction. The call in Section V-C for simulations at k = 12 and 15 “to locate the phase transition point more precisely” could note this.

· The analysis code and model are available only on request. Depositing them publicly, for example on Zenodo, would allow the results to be reproduced.

Competing interests. I have written a working paper on withdrawal-date coverage in Queued Up data for the non-ISO West (doi:10.5281/zenodo.22547678). I have no other competing interests.

Competing interests

I have written a working paper on withdrawal-date coverage in Queued Up data for the non-ISO West (doi:10.5281/zenodo.22547678). I have no other competing interests.

Use of Artificial Intelligence (AI)

The author declares that they did not use generative AI to come up with new ideas for their review.