Weak gravitational lensing has a parity check, and the check keeps coming back not-quite-clean. Lensing by scalar density perturbations produces, to leading order, only E-mode (curl-free) shear patterns: the B-mode (curl) component should vanish, contaminated only by tiny higher-order effects, source clustering, and redshift-dependent corrections, all well below current sensitivity. B-mode statistics are therefore used as the null test of every cosmic shear survey: detect B-modes and you have, by assumption, found your systematics, in PSF modeling, detector effects, astrometry, or blending. CFHTLenS reported significant B-modes whose origin was never fully traced; KiDS analyses found B-mode contamination requiring scale cuts and revised PSF treatments; DES and HSC pass their headline null tests but at thresholds set by, and absorbing, the known residuals.
The structural discomfort is twofold. Practically, shear B-modes at any level cap the credibility of the E-mode cosmology built on the same pipelines: the S8 measurements central to the growth tensions inherit whatever systematic floor the B-mode tests merely bound. Conceptually, the dismissal is assumption-laden: B-modes are treated as impossible signals because scalar lensing in the standard model cannot make them, so every detection is definitionally a systematic, an inference that forecloses discovery. Vector and tensor perturbations, intrinsic alignment physics beyond the standard models, and any non-standard gravitational contribution would all imprint genuine B-modes that current practice would file under instrument error.
The standing is a methodological tension sharpening into a physics question: as Euclid and Rubin shrink statistical errors tenfold, the B-mode budgets become the binding constraint, and the field must decide whether persistent cross-survey B-mode floors are shared systematics or a signal the null-test logic was built to ignore.