opt33, you bring up an important point of fact in your narrative:
Failures are far more information than passing results.
An attempt to prove an overclock is stable is like trying to prove a negative; it's an attempt to prove it won't fail. That's useless.
Finding the points of failure is the key point of information we all need in stabilizing a machine.
In your observation I see one row of data (the 4.8 Ghz), and two voltage points relative to known long term failure (in games and AVX testing).
If I were evaluating a system with these figures, what I'd want in addition are the minimal failure/success points in voltage, all but ignoring thermal limitations, for 4.9 Ghz, or as close to that as I can get even by extraordinary means (as in, possibly more than just multiplier changes, getting maybe 4.86 or 4.87, whatever I could get the chip to do).
What I'm looking for is a chart showing the curve trend above the target (4.8Ghz in your case), to see what the slope of that curve is. If the curve shows the chip can't operate for even a short, simple test at, say, 4.85 at about 1.31, I'd expect the system can't run long term at 4.8 at 1.31.
In other words, it helps to understand the trend of the chip's demands for voltage over the range of frequencies, at least one good notch above the target.
This does mean the highest targets are the toughest. We have to use the curve coming up toward the target, from below, and that's not nearly as helpful, but it can be used.
One can track the curve from, say, 4.2 through 4.8, then taking the slope op the last two, that is 4.6 and 4.7, lower the slope (towards flat, as in towards infinite voltage increase) about 5 to 10%, and project the 4.9 and 5.0 requirements, knowing the chip can't actually run that fast.
With that, one can see a little more clearly just how close to the minimal margin the chip is running. You're already saying that 1.31v is quite enough to survive gaming at 4.8Ghz, and if you're not thermally limited you might be able to see if 2 or 3 hundredths is even likely to be of help. If the curve shows there's no such margin to the next step up, and it would need 1.36 or more, 4.8Ghz just might be over your usage pattern's limit.
Put another way, you can find the trend at a lower speed. Set at 4.6 Ghz, for example, and find a voltage that passes the other tests you mentioned, but fails the AVX - just like the 4.8 Ghz setting. Let's say it turns out that 1.27 passes the test but fails AVX. Then we see that AVX passes at 1.31 at 4.6 Ghz. This suggests the margin required.
The margin is also a curve, meaning that the 0.04v difference at 4.6 (to get AVX to pass too) will be larger at 4.8, perhaps 0.05v or 0.06v, but you at least have some guide as to the relationship taken from a lower speed test.
It doesn't provide proof, just a guide, a measured guess at the chips personality.