Over 1993-2016, studies have shown that the observed global mean sea level (GMSL) budget is closed within the current data uncertainties. However, non-closure of the budget was recently reported when using Jason-3, Argo and GRACE/GRACE Follow-On data after 2016. This non-closure may result from errors in the data sets used to estimate the GMSL and its components. Here, we investigate possible sources of errors affecting Jason-3 and Argo data. Comparisons of Jason-3 GMSL trends with other altimetry missions show good agreement within 0.4 mm/yr over 2016-present. Besides, the wet tropospheric correction uncertainty from the Jason-3 radiometer contributes to up to 0.2 mm/yr. Therefore, altimetry alone cannot explain the misfit in the GMSL budget observed after 2016. Argo-based salinity products display strong discrepancies since 2016, attributed to instrumental problems and data editing issues. Reassessment of the sea level budget with the thermosteric component provides about 40% improvement in the budget closure.