How long can you trust a Starlink TLE? Sweep outputs bundle (paper-tle-divergence-atlas)

Sweep outputs underpinning the open-science paper How long can you trust a Starlink TLE? An empirical comparison of SGP4 and high-fidelity propagation against operator-updated truth across a megaconstellation. Contents (flat layout): the aggregated main NRLMSISE-00 sweep (all_runs.parquet, 24,641 pairs across 501 Starlink satellites over April 2026), the CdA × {0.8, 1.2} and maneuver-threshold sensitivity-subset frames, the analysis-summary JSON outputs (propagator wins, H3 regression, CdA sensitivity, maneuver-threshold sensitivity, truth-floor diagnostic), the mixed-effects supplement CSV, the sweep-statistics text report, the GMAT mission script, the resumable sweep manifest, and the EGM2008 installer. The companion code repository is at github.com/astro-tools/paper-tle-divergence-atlas.