500K One-Commit Global Impact

Amplifies the "exactly one commit" effect discovered in the small <150 sample in Document 14 directly onto the official full-scale conductor<500,000 dataset for measurement. The paper reports that out of 3,064,705 curves, 1,170,876 have analytic rank 0; narrowing down to semistable/optimal/conductor with at least two prime factors leaves 178,364 pre-candidates, and the current Algorithm 1 accepts 36,687 (about 20.5686%). Git compare shows that the same commit caused a net change of +2/-4064 in ec_labels_500k.txt; after deducting metadata lines, this exactly corresponds to 4062 curve rows being removed, reconstructing the old accepted count of 40,749—a removal rate of 9.9683%, dropping the pre-candidate acceptance rate from 22.8460% to 20.5686% (a drop of about 2.28 percentage points), with an impact on all curves of about 0.1325 percentage points. Because Algorithm 1 in this commit has only two theorem-level changes (strict $3/5/7$ isogeny exclusion + independent $a_3\ne\pm3$), every removed curve must fall into the union of these two—this is set-level closure, but the exact per-gate histogram at the 500K scale remains unknown, and the document explicitly warns against directly extrapolating the $9/2/1/1$ ratio from <150.