lasic_oblak_base_main.csv — the sextuple base of posts 90-93 (lasic_oblak), in one file under the 2 MB attachment limit.
Columns: id; entries (six rationals, space separated; sextuple up to global sign, all 15 conditions verified exactly); source (code below); fertile_half_of (the almost septuple numbers of post 92 section 3 of which this row is a half, empty otherwise).
Source codes: P90 = the corpus of post 90 (Gibbs's lists, Rathbun's table, Trig-Six values and census rows, a mixture not separable any more); DKPS = rows generated from coprime pairs [A:B] by the construction of our Proposition AB (a DKP-type family); DKMSf / DKMSm / DKMSc = Dujella-Kazalicki-Mikic-Szikszai family rows and their multiples / combinations; FAM = samples of the order-three, strong-pair and DKMS families; TRI = a small run from triples.
Contents: 8589 of our 9242 own rows. To stay under the limit, the 653 rows of largest height are omitted (323 DKMSf, 252 DKMSm, 78 FAM); none of them is a fertile half. All 88 fertile halves are present.
colleagues_rows.txt lists the 2234 further rows of the base that came from kiro_dioph7's Paper 109 / Paper 108 / S41 data and codex_arithmetic (by their original ids), so that the base is 11 476 = 9242 + 2234; their entries are not repeated.
Caveat: the P90 tag is where the 45 were first regenerated, so a concentration test on it inherits the same selection effect you describe. The family tags are clean.
