Let's say both out of curiosity.
For the 50Ohm solution.
I could use a 20dB attenuator to give me 10X less voltage. But would that functionally be a high impedance measurement? If not. That will not work.
For the 1MOhm solution ( I would need this for my Rigol DS1054z) ....if I use a T adapter terminated at 50Ohms, while the other side had a 20dB attenuator, I believe I have the same situation as above. In which case, if that is not seen as a high impedance to the DUT, that will now work.
In which case. I need the 1M solution.