One-Commit Semantic Autopsy

Directly answering the question left by document 13: the old fixture commit (1a0489c…) and the current commit (31fae20…) differ by exactly one commit, and the document dissects line by line what mathematical semantics this single commit actually changed. The change in Algorithm 1 is the most critical: the old version did not uniformly exclude $3,5,7$-isogenies, but established a conditional set $A_{\mathrm{old}}=\{3:3\mid N\text{ or }|a_3|=3\}\cup\{5:5\mid N\}\cup\{7:7\mid N\}$, only excluding rational $p$-isogenies for primes falling into this set; the new version changed to an unconditional $A_{\mathrm{new}}=\{3,5,7\}$, and additionally added the independent condition $a_3(E)\ne\pm3$—the document explicitly judges this as a substantive tightening of the theorem predicate, not a performance refactoring. The change in Algorithm 2 is more subtle: the same commit tightened $\gcd(M,N)=1$ to $\gcd(M,3N)=1$, but simultaneously removed the old version's twist-side disc_valuation_condition—so in Algorithm 2, a shrink mechanism and an expand mechanism exist simultaneously, and the effect of the entire commit cannot be simply summarized in the single direction of "the new version is stricter." This document truly advances the discipline of "reason=OPEN, no guessing" from document 08 a step further: instead of guessing the reason, it precisely pinpoints the single commit and dissects its mathematical consequences line by line.