If there are a lot of columns that could be processed separately, e.g. imputation, mapping values, etc, then it would be better to have separate nodes for these right? The upside is modularity (as you mentioned), which also means the nodes could be run in parallel. The downside is that there would be a lot of nodes;

process_col_A

,

process_col_B

, ...,

process_col_ZZ

.