CDO

Hello,
I'm new to CDO, got to know about it because DWD suggest it for use to convert ICON icosahedral grids to lat-long.
Compilation of cdo worked fine:

cdo -V
Climate Data Operators version 1.9.9 (https://mpimet.mpg.de/cdo)
System: x86_64-pc-linux-gnu
CXX Compiler: g++ -std=gnu++11 -g -O2 -fopenmp -pthread
CXX version : g++ (GCC) 4.8.5 20150623 (Red Hat 4.8.5-39)
C Compiler: gcc -std=gnu99 -g -O2 -fopenmp -pthread -pthread
C version : gcc (GCC) 4.8.5 20150623 (Red Hat 4.8.5-39)
F77 Compiler: gfortran -g -O2
F77 version : GNU Fortran (GCC) 4.8.5 20150623 (Red Hat 4.8.5-39)
Features: 62GB 40threads C++11 OpenMP3 Fortran PTHREADS NC4/HDF5 OPeNDAP SSE2
Libraries:
Filetypes: srv ext ieg grb1 grb2 nc1 nc2 nc4 nc4c
     CDI library version : 1.9.9
 cgribex library version : 1.9.5
 ecCodes library version : 2.16.0
  NetCDF library version : 4.3.3.1 of Dec 10 2015 16:44:18 $
    hdf5 library version : 1.8.12
    exse library version : 1.4.1
    FILE library version : 1.9.1

Using the target grid and weights provided by DWD at:
https://opendata.dwd.de/weather/lib/cdo/ICON_D2_002_EASY.tar.bz2

Conversion from icosahedral to lat-lon does work fine in principle.
E.g. for 2m temperature downloaded for 2 time steps:
https://opendata.dwd.de/weather/nwp/icon-d2/grib/00/t_2m/icon-d2_germany_icosahedral_single-level_2021012100_004_2d_t_2m.grib2.bz2
https://opendata.dwd.de/weather/nwp/icon-d2/grib/00/t_2m/icon-d2_germany_icosahedral_single-level_2021012100_005_2d_t_2m.grib2.bz2

cdo -f grb2 remap,target_grid_icon_d2_002.txt,weights_icon_d2_002.nc icon-d2_germany_icosahedral_single-level_2021012100_004_2d_t_2m.grib2 output_4.grib2
cdo -f grb2 remap,target_grid_icon_d2_002.txt,weights_icon_d2_002.nc icon-d2_germany_icosahedral_single-level_2021012100_005_2d_t_2m.grib2 output_5.grib2

However, while conversion of the +4h time step conserves the time information correctly in the output grib, time information in the +5h output is changed. Instead of forecast with run + 5h in the original file, time is coded as analysis at (run + 5h). So the timestamp is somehow conserved, but the meaning is considerably changed.

This is even more problematic for accumulated files like total precipitation.
E.g.:
https://opendata.dwd.de/weather/nwp/icon-d2/grib/00/tot_prec/icon-d2_germany_icosahedral_single-level_2021012100_004_2d_tot_prec.grib2.bz2 contains precipitation forecasts at +4 (as range 0-4), +4:15, +4:30 and +4:45. If converted as outlined above the time information is correctly translated for +4, +4:15 as run + forecast range, while changed to "analysis" for the other two fields.

Looking at a grib_dump nothing unusual shows up, only maybe that DWD is using minutes for the forecast range and not hours.

Is this a bug in cdo or am I doing something wrong?
Any help would be very much appreciated.

Regards,
Martin