The problem is bypassed by saving the dataframe with JLD2 rather than JLD and using JLD2 to open in the Pluto notebook.