Let's look at this phenomenon in a bit more detail. Look at the curves in Fig. 3.
The top chart in Fig. 3 is a representation of the M-MO curve shown previously and is limited to only showing the chromium-oxide curve from that chart. This chromium-oxide curve is the equilibrium curve at which the equation MO↔M+O exists for chromium. Thus, along any such curve, there might be equal probability for metal-oxides to break up into the pure metal plus liberated oxygen or the metal to react with oxygen to form that particular metal oxide. It could theoretically go either way. But as we move further to the right of that curve, the reaction becomes more strongly one of oxide reduction instead of oxide formation. And when we are at least one diagonal (the diagonal of one of the blocks) to the right of that oxide curve, as represented by curve B, there is a stronger and stronger probability for the reaction to go only one way, namely MO ↔M+O. The further to the right one goes, the stronger should be that dissociation/reduction reaction.
Now, let's look at the bottom chart in Fig. 1 (fromPart 1), showing several variations of a "Curve C." These curves represent the progressive oxidation that occurs to metals as they are being heated in an atmosphere, be it a gaseous atmosphere or a vacuum atmosphere (there are still lots of air molecules present in any vacuum-brazing cycle). If the atmosphere quality is poor for any reason, then the amount of oxidation that occurs in the heat-up portion of the brazing cycle might be quite significant, such as that suggested by the upper curve (C1) in that group. Note that as the heating approaches the theoretical equilibrium line for that particular oxide (chromium-oxide is used in this figure), the rate of oxidation slows down and then stops. As you progress to a higher temperature (for that particular given atmosphere condition, dewpoint, etc.) in that furnace run, you will note that you will be moving to the right of the equilibrium curve and there will be a driving force to dissociate those oxides that have formed. Please note that by the time curve C1 re-crosses the horizontal line (representing the surface condition of the part at the start of the brazing cycle), a much higher temperature is required than if the atmosphere quality (dewpoint, etc.) had been much better (such as that represented by the other curves, C2, C3 and C4).
From this lower chart it can be seen that the better the quality of the atmosphere, the less will be the amount of oxidation that occurs in the cycle and the easier it will be (temperature-wise) to reduce/dissociate those oxides.
Interesting visual evidence of all this was noted in a brazing cycle in which the furnace quality was somewhat poor, and the brazing temperature had to be quite high to get "one diagonal to the right" of the given curve. When the stainless component was removed from the furnace, it had an etched surface look to it instead of the bright shiny look it had when it was loaded into the furnace. An identical piece brazed in a high-quality atmosphere came out of the furnace still as bright and shiny as when it went in. Thus, an added quality-control item might be to look at the "degree of etching" you note on the surface of the stainless (or other metal) when it comes out of the furnace. That can be a direct indicator of what occurred inside the furnace chamber during the brazing run.