Differential expression in Scanpy

Hello! I am new to Scanpy and was looking into performing normalization and differential expression testing with scRNA-Seq data. From the tutorial, the usual count+log(1+p) normalization is used, before being used for DE testing.

However, if I were to use Pearson residual normalization, then do I have to eventually fall back to count+log(1+p) for DE testing? From what I understand, scanpy.experimental.pp.normalize_pearson_residuals returns Pearson normalised residuals. But from SCTransform v2 paper, it was stated that DE testing with residuals may lead to high false positives, and suggested using corrected UMI counts from the residuals to perform DE testing. So in this case, which is the best for DE testing? Thanks!