No luck yet. I tried to scale back the timings a bit from the settings I use with two DIMMs. As soon as I set the frequency greater than 2667 with four DIMMs installed the system refuses to even POST.
I think I am done playing with this for now. I updated my memory specs on the rebuild to read 16Gb at 3200 MHz. I also posted the final comparison in 3DMark between the new and original Monolith builds below.
Link to the final 3Dmark run!
AMD Radeon R9 Fury X video card benchmark result - AMD Ryzen 1800X,ASRock X370 Professional Gaming
Update to the build. With the UEFI version 2.36, I have been able to run all four DIMMs at the rated speed. I have added the CPUz information below.
It feels great! I feel that I now have the system in a good spot and can finally get on with enjoying the system rather than tweaking more UEFI settings. I did notice that ASRock released the final 2.4 UEFI today, but I'm not sure I want to upgrade, since everything is running fantastic on 2.36.
I did see a pretty big jump in speed using the Gears 4 benchmark. I like that benchmark as it displays CPU and CPU render fps regardless of any GPU bottleneck.
I also created this chart using the benchmark and different RAM speeds using two single rank DIMMs and a final test with all four installed at 3200 MHz. The difference between 4 and 2 DIMMs at 3200 surprised me. I think that may have something to do with the latency improvements introduced in the 1.0.04a AGESA (all my test other than the 32 GB were done with a 1.0.0.3 based UEFI)
From where I was sitting (2667 CL14) I have gained about 19% raw CPU performance in this title.
Looking at your Firestrike scores, it seems like the bulk of your gains were in the GPU side? It almost looks like Crossfire wasn't being correctly implemented in your original benchmarks, but now it certainly is. You do have a nice 14% gain in your physics score, likely from the 400MHz you were able to add to the 1800X! 
Pretty sweet system, I can only hit about 15000 for a total score with a single Fury X.
Woot! (Forgive me, but it seems appropriate)
Upgraded to the new AGESA 1.0.0.6 full release for my board. Made additional gains! Still learning about all the new settings available, perhaps I can squeeze out a bit more.
Really impressed with the platform updates since launch! Guess I'm not finished messing around with this after all. 
Thanks!
I actually have some further updates to the build. I have managed to get 3466 MHz stable in a 4X8Gb configuration. I needed to switch the memory command rate to 2T and set the ProcODT to 48 ohms. Once those changes were made, I've been able to pass HIC memtest over 450% so far.
