"Cold bug" to me means that a proc won't work at all under a relatively low -C, like -15C. I don't think anyone could call issues at colder than -100C a "cold bug"
I've said a lot of times that CedarMill cores seem to be 1) stable pretty close to their suicides for any given cooling and 2) don't respond drastically well to substantial decreases in temps. The latter is particularly true on Preslers. If they are pretty good on air or water, that's really the telltale sign if it's a good clocker at all.
For me stock cool air->cool water was 200-300MHz more. From cool water->phase was also only 200-300MHz and that was a huge drop in temps. From phase->DI (20-30C less than phase with a lot more capacity) also gave 200-300MHz, but that was a lot of work and trying for over an hour. Stable gains on DI were more like 150MH with a constant feeding of DI for even that. Even at idle, DI was not enough capacity to keep the temps from running away > 5.8GHz/1.70Vc with the proc *idling*. There's a very good reason any Presler screened >6GHz is pretty much guaranteed to be running on LN2
I really don't know what crime/fishy's issue was on the 631...I am still not 100% convinced it was a cold bug. I've run across issues that might mask themselves as that...like tightening the evap too much and bending the board just a bit too much. For a week until I figured it out, it would boot OK when "warmer", but after it idled a bit and cooled off more, it would lock up and then not reboot unless I let it sit for a long while. Turns out that board was bent "just enough" to be fine when warm, but when things cooled off more and contracted, it would cause a problem. I loosened the mounting screws 1-2 turns each and the issue magically disappeared. To this day, I tighten the evap until it won't POST and then back off 1-2 turns on each screw...that's how I know when it's tight enough, LOL.
Lastly, I've see really strange behavior on CM cores. Even ones that can do much better than mine on air don't do any or much better on water. I mean I've seen Preslers that will do upwards of 4.6-4.8GHz at default Vc and mine needed Vc starting around 4.2GHz IIRC. That clearly seems stronger than mine, yet they still cap out out in the same 5.0-5.2GHz range on water, just like mine.
I dunno, I still think it's the proc limits. All but the strongest CM cores seem to start running into trouble around 6GHz +/-200-300MHz regardless of cooling. Seems like 6.2GHz+ is limited to the best cores on the best cooling at that. That problem is compounded on Presler since BOTH cores need to be gems.
FSB is definitely an issue too for heat/stress. All the XEs that are much >6GHz are not only doing it on LN2, they are at extremely high multis (low FSBs). My guess is they can't possibly do it any other way...just too much stress.