Perhaps somewhat off-topic, and something along the liens of nit-picking, but closing up gaps generally has little to do with work hardening. The gaps normally found in the crystalline structure (due to the fact that spheres can't stack perfectly) can't really be closed up, and depleting the material of vacancies (spots in the grid where there should have been an atom, but isn't). Rather, crystalline materials (such as almost every metal in use) flow due to the generation and movement of faults in the crystalline matrix. Now, these faults can act as obstacles for each other, so as more and more are created, it becomes harder and harder for them to move, and thus the material becomes harder.
When we heat a material, we make it easier for the different faults to move around (the extra thermal energy can supply the extra "push" needed to overcome an obstacle). This can result in some of the worse "traffic jams" dissolving, and in some cases some faults may cancel each other out, come to the surface of the material, or otherwise disappear, making the material softer again. If the deformation was severe enough, then we might also get a recrystallisation, where the crystalline structure is basically recreated from scratch, overwriting the old one, and the faults in it, basically re-setting the work hardening to zero.