Extrapolators should make sure that at pipeline creation or at runtime, that the dimensions of the feature_input_variable and grid_variable match. Without this check, this can lead to a cryptic error about array sizes that doesn't hint at the solution.
This would apply to all classification extrapolators.