I can only support the criticism of the other commenters:
Usability has deteriorated by the new weather model selector and a number of new deficiencies been introduced

• selector can often not be closed when ECMWF is active
• selections are sometimes not effective (although this bug may predate the current implementation
• re-centering the display with new selection is a nuisance
• when selecting ICON-D2 erratic time jumps occur - first forward then backward
  which means that when toggling between ICON-D2 and other models the time steps back continuously

It seems this UI re-design was neither planned nor executed/QC'ed carefully.

I can only support the criticism of the other commenters:
Usability has deteriorated by the new weather model selector and a number of new deficiencies been introduced

• selector can often not be closed when ECMWF is active
• selections are sometimes not effective (although this bug may predate the current implementation
• re-centering the display with new selection is a nuisance
• when selecting ICON-D2 erratic time jumps occur - first forward then backward
  which means that when toggling between ICON-D2 and other models the time steps back continuously

It seems this UI re-design was neither planned nor executed/QC'ed carefully.

in the animated display of weather radar data one can observe distinct aliasing effects:
local, fast moving variations in precipitation are shown to move in directions opposite to larger scale aggregates, without physical indications like vertical variations of wind directions.

Most likely these are aliasing artifacts which do not occur in the animation of much coarser time series of satellite data. Obviously you are applying a powerful anti-aliasing filter to the latter - which should also be applied to weather radar data.