q8 — leading coefficient 1.1021 to 1.1020

q7 lifted the aspect-10 compact bound and replaced printed $1.1026$ by $1.1021$ at $n=33$ bins, target $0.9111$. The leftover slack was unused $\varphi$: SLSQP sits at $0.91122$. $R\le 9$ with the mass-opt cut cannot beat $1.1021$.

This folder keeps the split at aspect $10$ and raises the face target to $0.9112$ on the same $n=33$ matrix. Faces certify $Q\ge 0.907507$. Mass-stationarity still forces any used support of aspect $\ge 10$ to have $Q>10/11$. That replaces $1.1021$ in the same Section 7 chain.

The withdrawn $1.1168$ stays withdrawn. Finite $Z$ and $N_0(Z)-Z$ stay residue. $s>3$ along Lemma 4.3 stays residue. $n=34$ at this split is still a predicted $1.1017$ if faces certify.

./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_n33_t0p9112.txt (not re-enumerated on the fast path). Leftover probes in work/.