So, update from last night's testing before I gave up: Dipswitch reversion did not help.
I've now tried replicating the circumstances that led to the jump in performance, driver changes with DDU etc, did not work.
Tried different BIOS switch positions with shut downs in between, did not result in the performance increase.
Reverted dipswitches to stock, and much like when I changed them to begin with, zero effect noticeable.
The only GOOD thing so far is apparently +1000 vRAM is actually "stable" - it passed TimeSpy Extreme Stress Test with 98.9%, and Port Royal Stress Test with 99.4%, and it got hammered with 2 hours of Cyberpunk with no crashes & no visual issues as far as I noticed - this was with the visual settings all maxed, Ray-Tracing all on & the last setting on Psycho, with DLSS on Quality - so basically the game looking graphically as great as it can and hammering the GPU as much as possible, and it had no issues, despite running 2160MHz @ 1068mV (typically it was running at 2145-2160) & +1000 Memory, with no Classified Tool. I'm quite happy with that, especially considering the temps are still totally fine, highest GPU temp was 52C, Mems were in the 50.9C on the lowest to 65C on the highest, with a Mem Junction Temp of 72C peak (it was typically in 66C-70C range when I was looking at it, I have it in the OSD for MSI Afterburner via HWInfo64 now) & the VRMs peaked at 38C on the lowest & 54.5C on the highest. So considering it's running +120/+1000, the card is running pretty cool, hopefully those temps will go down even more once I get the next round of heatsinks on the back plate.
But, yeah, benchmarking performance is back to being stuck at around 14.9k max with +150/+1250 for no apparent reason.
If anyone has any suggestions for other things to try to see if I can get it back to where it should be, I'll give'em a try.