q4 — leading coefficient 1.1118 to 1.1057

q3 lifted the aspect-12 compact bound and replaced printed $1.1185$ by $1.1118$. The leftover slack was the face target and the mid-radius binning error. This folder raises $\varphi$ and $n$ on the same Rayleigh. The winning row is aspect $10$, $n=26$, $\varphi=0.9091$, so $Q\ge 0.904414$ on that class. Mass-stationarity forces any used support of aspect $\ge 10$ to have $Q>10/11$. Every finitely atomic radial measure reduces to one of those two classes. That replaces $1.1118$ in the same Section 7 chain.

The aspect-$\le 4$ number $1.1087$ is not used as a class-only quote: the unrestricted leading $1.1057$ sits below it. The withdrawn $1.1168$ stays withdrawn. Finite $Z$ and $N_0(Z)-Z$ stay residue.

./run_all.sh

Certificates: certs/lift.json, leading.json, aspect_identities.json, lift_stdlib.json, rebuild.json. Independent algebra: verify_aspect.c, verify_aspect.rs. Faces dump: certs/beta3_mid_faces_R10_n26_t0p9091.txt (not re-enumerated on the fast path). Second enum: certs/faces_rs.txt (Rust Cramer, copositive). Leftover probes in work/.