Far-end crosstalk (FEXT) can reduce the eye opening and eventually decrease the maximum data rate that can be transmitted through the high-speed interconnections. Therefore, FEXT is an important concern in high-speed digital design. The contributors of stripline FEXT include the dielectric inhomogeneity and the proximity effect. In addition, a characterization technique for the effective relative dielectric constant (Dk) of both core and prepreg has been proposed recently for the analysis of the inhomogeneous medium and induced FEXT. In our study, the FEXT levels of the striplines on various printed circuit boards (PCBs) are measured and compared. It is brought to our attention that for some PCB striplines sharing the same stack-up, same PCB material, and manufactured by the same vendor, the corresponding measured FEXT magnitudes of these coupled single-ended traces could vary drastically, which may bring great challenges to the hardware engineers about FEXT level control during the high-volume PCB production phase. In this paper, the root cause of this issue is investigated and analyzed. The 'critical resin content percent' concept is proposed to explain the variations in the Dk values of prepreg that result in the FEXT level variance. The full wave simulations are conducted to identify the 'critical glass weave'. A measurement-based statistical analysis is performed to verify the 'critical resin content percent' concept. A design guideline for FEXT control strategy in the high-volume PCB manufacturing is presented based upon this investigation.