Thanks for the response, and I'd certainly be happy to check it out. Optimization, testing, trial and error was not really the level of detail that I was hoping for though, in comparison to the observation that SnappyCam was using new DCTs specifically tailored to the co-processor instruction set. It's not that hard to speed up processing simply by discarding data.
Sapphire's high moduli and low strain to failure probably mean that it will never have that degree of flex. It can be produced in sheets - at least as large as an MBP screen - but it's expensive and such sheets are typically much thicker than GG (several mm). I could imagine a thin sapphire layer overlaid on GG for increased hardness, but their different coefficients of thermal expansivity might be an issue.