This applies to callbacks (not callback interfaces). Right now both unwrapping and wrapping assume nullability without failing codegen if it's not present. We should fix that.
Unwrapping got addressed in bug 792137. Not sure how much we care about wrapping. The obvious change to wrapping would be to use ObjectValue() when not nullable, but I'm not sure there's a win to be had there over using ObjectOrNullValue() no matter what.
This is fully fixed now, with a null-check on the CallbackFunction* and everything, in bug 779048.