For a few decades now it has been commonly believed that the near-lack of nova eruptions in cataclysmic variable (CV) systems with orbital periods in the range $~2-3$ hours may be explained by the donor red dwarf (RD) being eroded down to below roughly $\sim0.35M_\odot$, for which the star becomes fully convective, deeming magnetic braking as an inefficient angular momentum sink. Thus, the donor radius shrinks and Roche-lobe overflow (RLOF) resumes only after the orbital period emerges at the bottom of the period gap. However, this theory does not explain the existence of the occasional nova (or dwarf nova, or nova-like) detected within the period gap. The mechanism proposed here explains both the existence of the gap and of eruptions within the gap by using a self-consistent evolution code that accounts for both stellar components and their binary separation simultaneously. The method reveals that mass transfer resumes well before the orbital period shrinks below the gap, thus explaining the observed eruptions within the gap as well as exhibiting an orbital period distribution remarkably similar to that observed.