Repeating earthquake sequences on the creeping section of the San Andreas Fault near Parkfield, California, provide a unique opportunity to investigate fault slip processes. These events display unusually long recurrence intervals high inferred stress drops, raising questions about the fault mechanics that govern their behavior. Lui and Lapusta (2018) proposed numerical fault models that reproduce many observed characteristics of these repeaters and proposed two end-member triggering mechanisms: enhanced dynamic weakening through thermal pressurization (TP) and elevated normal stress (ENS). Here, we re-examine these repeaters by assessing the consistency between source parameters inferred from seismic observations and those produced by the numerical models. Using detailed slip histories from the numerical models, we generate synthetic far-field waveforms to estimate the source-time functions and seismological stress drops using standard spectral methods. TP models generally yield stress drops comparable to field observations, while ENS models produce lower but still observationally consistent values. In both cases, seismologically inferred stress drops exhibit greater variability than the corresponding moment-based values directly computed from the simulations, underscoring methodological sensitivity in source parameter estimation. In terms of rupture characteristics, ENS models tend to produce longer and more variable source durations than TP models. Ongoing analyses compare simulated and observed source-time functions (STFs) and explore sensitivity to physical parameters such as shear-wave velocity and normal-stress heterogeneity. Together, these results help constrain fault properties and weakening processes responsible for field observations and clarify how fault physics is expressed in seismic observations.