Curious coincidence, I have been working on similar stuff lately. I've been working on a project that needed injecting a considerable amount of code inside some functions in the JJ2 executable. I COULD have used the spare space at the end of the code section, but that space would have exhausted anyway in the end, jumping all over the place isn't healthy cache-wise, and the compiler used for JJ2 is apparently pretty crummy given the results; so I figured I could optimize the code myself by hand, compressing the existing functions and inserting my own stuff in it.
I'd say that over time I became rather proficient at it; like, with the keyboard state processing routine I managed to save about 47% of the space, while preserving original functionality (hopefully) and without hammering down performance too much.

Just saying that, if you need a hand with this kind of stuff in JJ2+, I'd be happy to contribute.

P.S.: Hi guys, long time no see.