I had a PM from someone who built the probe and then wondered why the output noise was so bad. They were measuring about 100mV p-p at the output with the inputs shorted together. At first I had a difficulty believing the noise was that bad, but I got the same result when performing the same measurement. In fact one of my probes was a bit worse than 100mV p-p. So why is it so bad -- I don't know the answer yet, but I can confirm that LTSpice predicts the same result.

To get an answer I ran a .Noise analysis on the entire probe (with parasitics, but that doesn't matter much). LTSpice predicts that total RMS noise voltage at the output is 2.1832mVRMS. I use a fat-finger rule to get p-p from RMS by multiplying the RMS value by 6, so peak-to-peak output noise is about 13.1mV p-p at the scope input, if it is measuring with 1X. If you select the 10X probe setting then the noise will jump to 131mV p-p, which is what I'm seeing. 

This doesn't bother me too much because most of what I measure with these probes is repetitive in nature and I can average the noise to a much lower value. But you might have a problem if you need to dig a small, non-repetitive signal out of this noise.