Just an update:
We found what causes this issue. Basically when the ds.cbind (or the ds.dataFrame) function is used across more than one datasources, the function initially gets the column names of the input components from each study separately, then generates a unique vector of the column names across all studies, and then passes this unique vector to each study to be used as the column names of the generated data frame in each study. However, when the order of the columns of the input components is different in different studies then the function assigns the column names by the order that is passed through the unique vector without considering the actual order of the variables across the different studies.
We have corrected this behaviour and now the new versions of ds.cbind and ds.dataFrame functions that will be included in the v6.0 release of DataSHIELD work independently for each study.
Also, we plan to develop for a future release a “rename” function to allow the user to rename the column names of a dataframe and a “reorder” function to allow the user to re-order the columns in data frames.