We revisit the conventional implementation of the determination of $V_{us}$ via flavor-breaking (FB) finite-energy sum rule (FESR) analyses of inclusive hadronic $\tau$ decay data, which is
known to produce results $>3\sigma$ low compared to determinations from kaon physics and the expectations of three-family unitarity. We show that this implementation fails self-consistency tests, and that the source of this problem is a breakdown of assumptions concerning the treatment of higher dimension OPE contributions. We then provide an alternate implementation of the FB FESR approach which cures these problems. Lattice data for the relevant flavor-breaking correlator combination is also employed to clarify the treatment of the slowly-converging dimension $2$ OPE contribution to the relevant sum rules and quantify the associated truncation uncertainty. We implement this new approach using ALEPH non-strange data, and a combination of ALEPH, BaBar and Belle strange $\tau$ decay data. Normalizing the exclusive $\tau\rightarrow K^-\pi^0\nu_\tau$ mode component of the inclusive strange decay distribution using the recent preliminary BaBar result for the corresponding branching fraction we find a result, $V_{us}=0.2228(23)_{exp}(6)_{th}$, in excellent agreement with the results of $K_{\ell 3}$-based analyses, and in agreement within errors with three-family-unitarity expectations, thus resolving the long-standing inclusive $\tau$ $V_{us}$ puzzle.