For what it's worth, and I can't presently provide any references to back any of this up, but it has been my experience (as a scientific glassblower) that the oxidation state of specifically the transition metals ('TMs' which are 'network formers') is influenced only by thermal conditions within the *body* of the mass, and is effected by free oxygen (O) only at the surface.
By 'network former' is meant that the transition metal oxide (TMO) bonds into the SiO2 network -- in effect contributing its O(s) in a covalent bridge to adjacent (eg. Si) metal network constituent(s). Depending upon whether the TM also *accepts* linking O bond(s) to some other adjacent metal matrix constituent(s), the oxidative state of any particular TM 'center' in question may alter. Although I have essentially no experience in 'art glass', it is my impression that this is precisely the phenomenon which is occuring in the 'striking' of colors. The phenomenon of 'color' describes the absorbative characteristics of the (outer) valence electrons of a material - a characteristic which is easily changed by altering oxidation states.
At the surface however, the situation becomes considerably more complex as the role of gaseous oxygen (whether neutral O2 gas or a neutral atomic plasma from a flame) comes into play. Additionally, there is the factor of 'bonding constraint', a surface phenomenon, which for now can be ignored. To appreciate what is occuring at the surface, it must be understood that heat from a torch is conveyed to the glass by convection, and is distributed throughout the mass of the glass by conduction. Although some heat is transferred by radiation, there is likely more heat radiated _from_ than _into_ the glass. The point here is that during heating there is a temperature *and* viscosity _gradient_ in the glass, with the exposed surface much hotter and less viscous than the body of the glass. As torch heating generally continues until the bulk of the piece reaches a lower, plastic viscosity, the temperature of the outer few microns of glass may become so elevated that the vapor pressure of this layer of glass reaches 1 atm (in other words, the glass will 'boil' or volitilize). Dissociation of higher VP components will of course precede this. What this in effect means is that the covalent bonds of the network are now so transient that they have essentially ceased to exist, and the structural integrity of the glass is defined solely by weaker ionic and van der Waal forces. In this environment, the TMO components of a glass may either be oxidized (gain O from surrounding constituents) or reduced (give up O to surrounding constituents). When occurring in the bulk of the glass, this may be seen as a color change, but because of the presence of excess, highly reactive O radical (neutral plasma) in the flame, the chemistry at the surface will be very different.
An example is Corning 0120 lead glass: Although it is a 'high alpha' glass, it must be worked in a hissing, oxidative flame to avoid the reduction of metallic Pb at the surface. This is *only* a surface phenomenon, and it is reversible and controllable.
That said, the point I wish to make is that it may be a mistake to assume that there is any 'oxygen' in a 'free' sense, that enters into any chemical reactions in the bulk of a piece of glass. O is an element which at lower temperatures forms covalent and ionic bridges between network constituent (eg. TMOs) and matrix modifying (eg. an alkali metal and to a lesser extent its oxide) constituents. At elevated temperatures these bonds may rearrange and the oxidative states of constituents may alter, but again there is never really any 'free' O involved. At the surface the presence of O may come into play, but more in the formation of coatings, most of which are invisible save for a few exceptions. As a final consideration, there are bonding constraints imposed by the assymetry of a surface. As the coordination of 'surface' glass is different from that of 'bulk' glass, there are very different physical and chemical properties exhibited by glass surfaces. For one thing, the viscosity of a surface is much lower than that of the bulk at any given temperature. For another, there is a considerable amount of singly-bonded O at a glass surface meaning that dry, clean glass presents a strong negative charge. Because of this surface charge, a glass surface is quite hygroscopic, and rapidly forms a strongly-bound hydrated layer which may actually serve to stabilize the glass surface from chemical attack. In summary, there is essentially no 'oxygen' in glass, only 'oxides', whose coordination changes and thus 'color' and physical characteristics are temperature dependent.
Last edited 02-18-03