FFT was written in C. Release date is 1997 (there was nothing better then) and towards the end of BATTLE.BIN, you see non-ASM ASCII comments detailing files (.C) and folders. I also found stuff about data heaps (which apparently the coder must explicitly mention in C). I'm sure stuff like this exists in other files too and I'm sure anyone with a hex editor can do this to find out more about the large-scale file organization of any other important file (WORLD.BIN, SCUS...).
As for code that's never checked in FFT? We already know that DEBUG mode and Sound Novels are in the English ISO, but not called during normal game procedures. I strongly suspect there's stuff like this in BATTLE.BIN, especially on some of the longer routines I've seen (like knockback).
I use pSXrel, WinHex or HxD (depending on the computer), as well as the tools downloaded from this site. I didn't need Renegade up to now, but I might get it to type up the hex for my massive formula hack.
At this point, I'm not exactly sure what other tools I need, though a map of PSX data sections would be nice (e.g. 0x80000000 is ROM, 0xA0000000 is something to do with the system [RNG is called there, as is system time?]...)
Reverse Engineering Source Code...you mean turning Hex back into C? I know there's commercial programs that do stuff like that (I have some freeware on my laptop that converted EXEs to C#/Java), but it's not 100% and in general, there's no perfect re-engineering software algorithm that can work on any piece of ASM, even for hex code that's simply a compilation of a C program (and even if it did, there's no guarantee the resulting program would be readable or efficient). Considering also that I intend to ASM hack a lot of the resulting code while under space constraints (and most compilers from any language to ASM use space inefficient algorithms because most of those tend to optimize runtime and minimize the chance for variable misuse/error), it'll be likely that any ASM code I write will yield gibberish when fed through a cheap reverse compiler (because I don't follow conventions that restrict "free" register use).
As for the duplication of effort, that's one of the major barriers in this project (the other being not enough ASM hackers). I don't know how to split the work that doesn't either result in that or someone doing boring grunt work.