We examine fossil foraminiferal assemblages from 20 sediment cores to assess sudden relative sea‐ level (RSL) changes across three mud‐over‐peat contacts at three salt marshes in northern Humboldt Bay, California (~44.8°N, −124.2°W). We use a validated foraminiferal‐based Bayesian transfer function to evaluate the variability of subsidence stratigraphy at a range of 30‐6000 m across an estuary. We use the consistency in RSL reconstructions to support estimates of coseismic subsidence from megathrust earthquakes. To assess the variability of subsidence estimates, we analyzed: nine examples of the 1700 CE earthquake (average of 0.64 ±0.14 m subsidence; range of 0.24 ± 0.27 to 1.00 ± 0.44 m), five examples of the ca. 875 cal a BP earthquake (average of 0.43 ±0.16 m; range of 0.41 ± 0.36 to 0.48 ± 0.39 m), and six examples of the ca. 1120 cal a BP earthquake (average of 0.70±0.18 m; range of 0.47 ± 0.36 to 0.80 ± 0.49 m). Our subsidence estimates suggest ~±0.3 m of within‐site (intrasite) variability, which is consistent with previous research. We also identify inconsistencies between sites (intersite) at northern Humboldt Bay greater than one‐sigma uncertainties, driven by variable foraminiferal assemblages in the mud overlying the 1700 CE subsidence contact. Therefore, we recommend at least two quantitative microfossil reconstructions across the same stratigraphic sequence from different marsh sites within an estuary to account for estimate variability and provide increased confidence in vertical coseismic deformation estimates. Our results have broad implications for quantitative, microfossil‐based reconstructions of coseismic subsidence at temperate coastlines globally. © 2022 The Authors. Journal of Quaternary Science Published by John Wiley & Sons Ltd.