Error handling in stacked learner PipeOpLearnerCV

[GeorBe]

I have stacked learners ( LearnerClassifAvg, LearnerRegrAvg) using crossvalidation. In some simulation replicates the number of folds set when the learners are created appear to lead to error (train error). However I do not find inbuilt error handling as can be done with individual learners (shown in the MLR3 book 5.2 Error Handling). I would like to be able to use encapsulation/fallback methods. Actually the ideal would be to encapsulate, retrieve error message AND switch to lower number of folds as fallback.
Can error handling also be added to learners derived from pipeline operations? Can error handling be optimized to with respect to choice of fallback learners?

Thank you!