When NightWolve was packaging his Xak3 and Ys4 patches a few years ago, he used PPF. I'm pretty sure that Playstation and Saturn patches are also distributed as PPFs.
IPS has a limit of 16MB addressability; otherwise it would probably be used for an image this large too (it's the most common for smaller ROMs).
For this reason (and a few others), Derrick Sobodash (aka 'D', 'D-Boy') came up with a universal patching format called 'Ninja'. But I haven't seen anything packaged with it.
I believe that "xdelta" and maybe even "bps" have pretty much taken over the world of binary patching since then.
You can even see BurntLasagna asking about using xdelta for the Ys4 patch back in 2012 ...
http://www.romhacking.net/forum/index.php?topic=15065.0I really like that xdelta is standardized and available on all platforms.
I don't recall whether PC-FX is based on ISO-9660 or whether it's raw data like PC Engine, but...
CUE/BIN patching has the additional drawback that all the ECC code in the sectors (304 bytes on top of a 2048-byte sector) is recalculated, and would need to be part of the patch. That's even for sectors with only a few bytes altered. But on the other hand, extracting individual tracks is more problematic (fewer people are able to manage, and there is sometimes problem data at the interface between tracks).
I definitely want to patch against the extracted tracks for exactly that reason ... but, as you say, getting those extracted tracks is often not as easy as getting a .cue/.bin image.
For instance, ImgBurn will natively produce a .cue/.bin and not extract the individual tracks, and I've only seen online rips of the Zeroigar CD in .cue/.bin format.
So I just extract the separate .cue/.wav/.iso tracks from the .cue/.bin before patching. It's pretty easy with BinChunker, so I built that into the patching process.
The PC-FX is raw data, just like the PCE, and not ISO-9660.
In summary, even though your patch will be large (much larger than you think it ought to be), I think PPF is the way to go.
Now that we're copying the original video instead of recompressing it, I've just tried the xdelta process again, and the xdelta patch is now down to 14MB in size.
That includes expanding the .iso file another 11MB to cope with relocating one of the videos to the end of the CD image.
That's something the only a delta-based patcher can handle with any kind of efficiency.
PPFStudio refuses to even try to create a patch with the differently-sized .iso files.