Currently, we raise an error:

ValueError: Pandas output does not support sparse data.

when a transformer does output sparse data and we try to wrap it inside a dataframe due to set_output(transform="pandas"). I assume that at this point, we could be extra nice and check if the transformer has a sparse_output keyword and advice changing it to False. We can also document that doing this change would be memory inefficient (it might blow up your RAM) but useful when it comes to data introspection.

What do you think about this proposal @thomasjpfan?