Quantized Tensor Train Compression in Fixed-Resolution Turbulence Simulations: Reanalysis of a Capped Reynolds Sweep and Criteria for Rank-Scaling Claims

Version 3 correction. This version supersedes the scientific interpretation in v2 (DOI: 10.5281/zenodo.18499334). No new simulations or raw-data validation are claimed. Version 2 interpreted a fixed-resolution Reynolds sweep under a maximum QTT bond dimension of 64 as evidence for Reynolds-independent rank, fitted an apparent zero rank-scaling exponent, inferred O(1) QTT degrees-of-freedom scaling with Reynolds number, and suggested implications for asymptotic DNS complexity. Those inferences are withdrawn. The published v2 table uses a fixed 128³ grid and Reynolds labels 50, 100, 200, 400, and 800. Every reported maximum rank is 64, equal to the imposed cap. Equality with a cap can be consistent with either adequate or inadequate approximation; without a stated fidelity criterion the summaries do not identify the required rank. The reported energy-balance values rise from 1.01 to 1.68, but their definition is insufficient to calibrate solution error or establish equal fidelity. Fixed resolution also does not determine what rank is required as finer physical structure is resolved. Version 3 retains the architecture description and numerical summaries as archival reports rather than independently validated benchmarks. It withdraws the unsupported asymptotic claims, separates core payload from peak working memory and runtime, and specifies a protocol for a future scaling study. The protocol distinguishes an ideal minimum rank from an algorithmically achieved sufficient rank; requires resolution-converged reference observables, stated tolerances and Reynolds histories; and separates snapshot compression from compressed time evolution. The source bundle contains the manuscript, bibliographic records, archival table data, plotting code, build instructions, and document-level verification tests. It does not contain the original solver or raw simulation evidence. Regenerating the figures and PDF is not reproducing the original turbulence experiments. The companion storage analysis is now available as Ceiling-Normalized Storage Bounds for Binary Tensor Trains: Exact scaling laws and compression crossovers for quantics representations, version 1.0.0, Zenodo, DOI: 10.5281/zenodo.22734260. Its storage results are separate from physical fidelity and do not validate the historical turbulence runs.