I think we don’t want to lose that capability at least for the expert interface.
I have never thoroughly explored a solution, but I believe that with the new standard, the stride may be read back into Fortran from a C routine that calls a C-Fortran descriptor on the input array: