Hats off to you, Bethesda. I'm astonished and at the same time happy that the problems concerning the generated game code are being looked into and also addressed!
^^ If that's going to have the same impact as that "unofficial speed patch", but to a degree where as it's actually been compiled that way it sorts out *EVERY* slow call, then this alone is going to make a massive performance boost. The upcoming changelog looks really good - much better than I had been expecting, I didn't expect to see any quest fixes for another couple of patches, or any engine improvements.
It could be done by using link-time code generation (/GL /LTCG) because the game consists of multiple libraries linked together: Usually, by the time the linker is invoked, the code will already have been generated and nothing more than dead references can be eliminated because the knowledge about how a libraries' functions work interally with the data passed to them is no longer available. By generating it at link time, the compiler essentially gains an overview of how each part of the program is related to another, enabling optimizations like automated inlining of functions in loops and where the speed/size tradeoff is favorable, partial unrolling to lessen the strain on the CPU's branch predictor for loops with small bodies, determining and elimination of code paths, function arguments and return values as well as global variables that normally would've been exported and thus linked into the final binary, enhanced awareness of alignment needs for data so other code won't be slowed down by attempting to do loads and stores on misaligned addresses, folding of multiple function calls for constant and pure functions (i.e. functions which don't have any side-effects) and many more.
Even if this wasn't used, these same optimizations are still applied at a library-local level and it'll also improve much of the obsolete code like using the much slower FSTSW -> TEST -> JNP chains instead of the equivalent floating point comparison instructions available on all processors since the Intel Pentium II. It'll probably will make a noticeable difference due to how many critical paths currently look like this or worse.