How experminetal is the experimental API?


I have been using scanpy for a while, and its just… awesome! thanks!
I have recently downlaoded a new dataset, and before re-running the same preproc recipe I was sing so far, I double-check for new/updated options.
I read about the Pearson residuals normalization, but just saw its on the experimental API.

Thus, how experimental is that implementation? Safe to go but not so-documentaed to be merged in the latest version?
As a bonus question, anyone could point me to which differentially expression framework is the most compatible one for Anndata/scanpy data? diffxpy has not been updated for a while…