I finally get to post some results! I am running VCSSA at 1.05v, so I didn't have to go to 1.10v for these modules to run at this speed.
I've gotten 32Gb (8x4Gb) of the Samsung 30nm modules stable at DDR3 1866 9-9-9-28 1T with 1.55v:
Have you tried the same timings and clock speed with 4x4GiB? If so, how much DRAMV did you need? I know I need less than 1.35DRAMV and 1.05VCCIO to get 4x4GiB stable with your primary timings, and clock speed. So I thought it was strange you needed 1.55DRAMV, over more VCCIO.