Oh well, that could be an issue. I recently updated SWATplusR to R 4.x. As there were some major updates in some of the R package dependencies I am not sure about the downward compatibility to R3.x.
Concerning multivariate calibration, in theory yes! SWATplusR just handles the communication between your SWAT TxtInOut folder and your R instance, takes care of parallel computing and tidying outputs ect. The simulation workflow that you build around that can be whatever you plan to do. This requires however that the user has to think about writing the analysys workflow and it is not just clicking through as it might be in SWAT-CUP or JAMES+.
Cheers
Christoph