PipeOpMice return m imputed datasets

[a-humble-learner]

Is there a way to set return_one to FALSE in PipeOpMice, so that I get m different imputed datasets? And then how do I feed each imputed dataset into a learner to train m separate models in the usual mlr3 piping language, e.g. po_mice %>>% po_classification_learner? And after that how do I pool the models, like that provided by mice::pool?