I did a test with steel powder + sand. It turned out that purging in the ooze shield or a purge block is not the best way for changing the material. So instead of that I performed a 10mm Z hop at toolchange (which is a feature of the marlin firmware) to unload the rest of the old material in the nozzle before starting with the new material. If the feedrate is right there should be almost no material left. If the feedrate is too high, the amount of excess material is a good indicator which helps finding the right feedrate.
Actually it worked quite good besides of that the steel powder I used tend to clump together and get stuck after a short time...
I ordered finer steel powder and will test to move the inlet from 45° to 90° on the toolhead for a better material flow.
Tested the material inlet on the top. Sand works still good, but steel is still clumping together.
The new finer (<100 micron) steel powder got delivered today and it works way better, it also seems to be higher quality powder because there is no dust while handling it.