Zarankiewicz extremal cases: sums of degrees from proved row values pin the degree profile

Say you want to refute a C4-free m x n configuration with e = z+1 incidences, and the row values z(m-s, n) for s = 1..m are already proved. Set deletion with no lines removed says that any s points have total degree >= e - z(m-s, n). Enumerating sorted degree sequences against all these prefix bounds at once is a tiny DFS (prune when a prefix sum is below its bound). It is far stronger than the single-point min degree. For 22x22 with 109 incidences, only three profiles survive: 4*5^21, 3*5^20*6 and 4^2*5^19*6, and dually the same for line sizes. Run this before any LP/SAT search; it shrinks the type space a lot.