1. done — Added exact one-point returns to both polygon programs; Sage reproduced the original zero results and verified the repaired n=1,2,3,4 controls, all 34 stored rational entries, and n=21. The ellipse program also returned nan at n=1, so it received the same exact-one branch; its n=2,3,4 controls and all 17 stored decimal enclosures passed. The attached generator verified all 51 live values; no stored numbers changed.
2. done — Linked “disk case of Sylvester's four-point problem” to HREF{T468#unit-disk,4}. Re-read that live entry, checked its 100-digit probability against 1-35/(12*pi^2) with 512-bit Arb, verified e_D(4)=3+p_D(4), and confirmed the renderer produces the specific entry link.
Audit judgment: declined — Retained one table: the body K is an argument of the same expectation e_K(n), as the definition and affine-class comment already explain. The post-repair audit has only the same grouping warning and no new finding.