Scikit-learn - Warm reloads for cross-validation

Question

In the developer's guide in the sklearn documentation (link below) there is a section heading that caught my eye - "An example of an algorithmic trick: warm reloads for cross-checking"

Unfortunately, at the moment, the section simply says: "TODO: demonstrate warm reload tricks to cross-check linear regression with reduced coordinates."

grid_search.GridSearchCV and grid_search.RandomizedSearchCV do not have a warm start parameter, so I assume that using warm restart during cross validation will require an estimate that supports warm start. But there is also an open github problem where the developer says, "Because of how GridSearchCV is developed (using clone, parallelism with n_jobs), setting the warm_start parameter will not be" It would be nice to create a mesh search object that could benefit from a warm start. "

Is there a way to implement the “algorithmic trick: warm reloads for cross validation” in sci-kit learn 0.17?

EDIT: or at 0.18 now that it is released?