We revisit the calculation of the strong couplings $D^{*}{D}^{*}\pi$ and $B^{*}{B}^{*}\pi$ from the light-cone sum rules (LCSR) using the pion light-cone distribution amplitudes. The accuracy of the underlying correlation function is upgraded by establishing the hard-collinear factorization formula at the leading power up to the next-to-leading order in ${\alpha }_s$. Furthermore, the next-to-leading power contributions are incorporated at the leading order by including the two-particle and three-particle higher-twist pion distribution amplitudes up to twist-4 accuracy. We perform a comprehensive numerical analysis with a systematic evaluation of theoretical uncertainties, predicting $g_{D^{*}{D}^{*}\pi }={5.71}_{-0.89}^{+0.70}{\mathrm{\text{GeV}}}^{-1}$ and $g_{B^{*}{B}^{*}\pi }={4.90}_{-0.67}^{+0.59}{\mathrm{\text{GeV}}}^{-1}$. Finally, by parameterizing the $1/m_H$ (with $H=D,B$) power corrections to extract the universal static coupling $\hat{g}=0.30\pm 0.04$, we compare our results with previous theoretical determinations and experimental data, and discuss the current limitations in disentangling the heavy quark spin symmetry breaking effects.