Do we care about k-fold cross validation for large datasets?

When considering incremental models how much do we care about k-fold cross validation? My understanding is that the k-fold bit is an effort to train on all of our data, while still protecting testing data for each fold. I wonder how often using all of the data is a requirement in the cases where `Incremental` will be used. My guess would be that this becomes less important.

The issue is an open design question about k-fold cross-validation for incremental models and large datasets. No files, tests, entry points, or acceptance criteria are named; first establish whether this support is needed and agree on the scope and success criteria before implementation.

Written by the indexing model from the issue text.