An accurate covariance matrix is crucial for reliable cosmological results in studying cosmic shear power spectra. By analyzing pseudo-$C_\ell$ estimates with different sky masks, researchers found that extreme sky cuts lead to increased off-diagonal covariance. Neglecting non-Gaussian contributions can underestimate uncertainties by up to 70%, with super-sample covariance being the dominant factor. Marginalizing over nuisance parameters in real cosmological analysis can reduce the importance of these covariance components.