See in "Original" to see the noise pattern clearly.

Assuming that ISO 50 is the same as ISO 100 with 1 EV overexposure and again pulled back I made a new noise test camparing ISO 50 with ISO 100, 200, 400, 800 and 1600 overexposed 1 EV and pulled back in Camera Raw 6.3. The noise level of ISO 50 and ISO 100 +1EV is the same. However, the highlight clipping is much larger for the ISO 50, which can be seen looking at the knob. The clipping in the black level is very close. I have no explanation for this - I have read on the web that the histogram should be exactly the same. But if these observations are true I will start using ISO 100 and deliberately overexpose.
I wouldn't hesitate a second using ISO 400 if I would want a faster shutter speed - the increase in noise is very limited. This is no surprise, however, this was not the practice for me in the field.