Motivated by the observed ~30% variations in flux from the L7 dwarf VHS 1256 b, we subjected its time-resolved Hubble Space Telescope (HST) WFC3 spectra (measured in two epochs: 2018 and 2020), as well as medium-resolution Very Large Telescope (VLT) X-shooter and Early Release Science James Webb Space Telescope (JWST) spectra to a suite of both standard Bayesian (nested sampling) and machine-learning (random forest) retrievals. We find that both HST and VLT data require vertically varying abundance profiles of water in order to model the spectra accurately. Despite the large flux variations observed in the HST data, the temporal variability cannot be attributed to a single varying atmospheric property. The retrieved atmospheric quantities are consistent with being invariant across time. However, we find that model grids offer relatively poor fits to the measured HST spectra and are unsuitable for quantifying the temporal variability of atmospheric properties. Additionally, our analysis of JWST spectra using model grids indicates consistency in terms of the retrieved properties across different wavelength channels. Despite the temporal variability in flux, the retrieved properties between HST and VLT, as well those as between HST and JWST, are consistent within the respective posterior uncertainties. Such an outcome bodes well for future retrieval analyses of exoplanetary atmospheres, which are expected to exhibit weaker flux variations.
