Thanks for your thoughts.
What you’ve described is pretty much what it is I think. The issue is with some of the remenant lines in the flare (the cooling section if I understand you) that are not atmospheric.
This isn’t a particulalry good graphic but if you look closely at the strong line at 777.4 nm, immediately to the left of it is another line. This can be seen if you examine the composite closely. (Purple is the flare section, blue is the weaker “gap” sepctrum and red is the fragment.)
This MAY be a triply ionised silicon line. In the only other case that I’ve recorded this type of behaviour, a Perseid fireball in 2013, this and other features were attributed by Jiri Borovicka as being deep blue lines overlapping from the next order. Jiri kindly provided a fully annotated graphic that I included in the Perseid short paper mentioned. Due to the quality of the Persied spectrum I couldn’t really argue (and of course it may very well have been completely different) but in this case I don’t think there is any doubt. Looking at the blue lines they suffer rapid de-focus and the same effect would be seen from the next order up. In this case the line is quite sharp so I believe it is a geniue emission at that wavelength. Also the other peaks in the purple line may be from silicon. Absolute identification is impossible, though due to the low resolution and that is the big problem!
There must be a short interval of time, of the order of a few milliseconds when the crushing pressure on the meteorid is a max and so probably is the temperature. This may momentarliy cause what might be called a “flash boiling” of the surface. The calculations are fraught with so many assupmtions that it’s impossible to say for sure what any given case is though. Clearly there is a fragment that exists for a few more milliseconds so it mast have had some material integrity. So what I think is being seen is emission from the cloud of meteoroid vapour generated at the max P and Max T if the meteoroid is tough enough to survive the whack it gets at this point. However if these lines are indeed silicon from the breakdown of SiO2 (silica/quartz) one of the peaks in the flare spectrum should (perhaps could ?) be O at 615.7nm and be of meteoroidal origin and not atmospheric.
I can’t lay my hands on it as I type but Jiri Borovika also wrote a paper in 1994 about a “twin temperature” model that could be used to explain particular emissions. I think these were in the region of 3000K and 10000K if memory serves. I’m afraid I just don’t know enough about the chemistry/physics/quantum mechanics to fully understand how it actually works. The best I can do is to read it as some “part” of the meteor is at one temp, maybe this is the hotter flash cloud. The another “part”, the surviving fragments, are at the lower temp due to whatever particular bulk thermal properties the meteoroid has.
We need more examples and better resolution!