Shortening timestep causes duplicate outputs

[n-beech]

Rerunning a simulation with a shortened timestep creates a duplicate output of the first timestep after the restart.

I think this is because the i/o counter is defined by finding the last record in the output file with a timestamp earlier than the current time and adding 1. If you need to rerun part of a simulation with a shorter timestep than was previously used (for example, after a blowup), the first timestep after the restart will not be overwritten because it will have a slightly earlier timestamp than the first output of the restarted run (old timestep - new timestep). You then have two near-duplicate records.

From io_meandata.F90:

               !_______________________________________________________________
                ! if the time rtime at the rec_count is larger than ctime we
                ! look for the closest record with the timestamp less than ctime
                do k=entry%rec_count, 1, -1
                    !PS if (partit%flag_debug)  print *, achar(27)//'[33m'//' -I/O-> call assert_nf B'//achar(27)//'[0m'//',  k=',k, ', rootpart=', entry%root_rank
                    ! determine rtime from exiting file
                    call assert_nf( nf_get_vara_double(entry%ncid, entry%tID, k, 1, rtime, 1), __LINE__)
                    if (ctime > rtime) then
                        entry%rec_count=k+1
                        exit ! a proper rec_count detected, exit the loop
                    end if
                    if (k==1) then
                        write(*,*) 'I/O '//trim(entry%name)//' WARNING: the existing output file will be overwritten'//'; ', entry%rec_count, ' records in the file;'
                        entry%rec_count=1
                        exit ! no appropriate rec_count detected
                    end if
                end do

Could the i/o counter be initialized by finding the number of records in the output file with timestamps earlier than the restart time and updated by 1 after each output?