Good eye, Scott.
Not only that, but the panel is unrestrained in the unsealed test whereas it is restrained in all the sealed tests. It looks as if the side supports are restrained as well - at least in the sealed tests, so that the restrained panel would act more as a membrane rather than a simple beam - that may be why there is point failure in the latter panels.
Interesting images none the less - especially showing how well the glass is adhered for the sealed situation.
As is it so difficult to lay glass on wet surfaces, I wonder if rough sanding is another method to keep the adherence up. Say 40 grit, heh heh.