I think I've found the problem.
I thought to myself, 2-2-2-5 always works, no matter what. Let's try 240 x 10 and memory 2.5-2-2-5 5/6 2.8v for 200MHz. This should still work just fine, only increasing the CAS latency while running at stock 200MHz with extra juice.
Rebooted, upon Windows loading up, immediate BSOD. Rebooted, and after post, it told me one of my Windows files was missing/corrupt. Rebooted, and couldn't get into the bios to change back to 2. Reset the CMOS jumper, got into the bios, changed back to 2-2-2-5 207MHz 2.8v. Running and stable again.
Looks like this memory is somewhat damaged, as only cas2 works, anything higher = unstable as hell.
But this leads me to believe ..... how could memtest86 pass @ 2.5-3-3-7 for so long? Odd, but looks like RMA to me. Thanks a ton to all who have helped.