Ok, this has me interested too. pullmoll moved most of the simulation into pasm, but one thing that it needed was a big hub buffer for the video. Move the display out to a touchscreen (which *can* do a 80x40 display) and you save all that hub ram. And if you drop in Kye's SD driver it works out of the box for pretty much every SD card I have tested.
No need necessarily for a Z80 emulation. 8080 was just fine (my preferred CP/M source was Z80 opcodes but only using 8080 ones so essentially 8080 code which is much simpler than Z80, ie Zicog). And if you only use 8080 opcodes we are back to Zicog which is simpler to run.
Keeping with the CP/M philosophy, one always rewrote the disk drivers anyway. Sorry I can't post all mine, they are custom coded for the N8VEM. Head to Peter Schorn's site as heater suggests
I like what you are planning. A few weeks back I finally cracked the CP/M file system code. You probably have too. And it isn't what is in the official textbooks. Basically, the file table is a list of pointers to the data blocks, and if the file is too big, put in multiple file table entries with the same file name.
If you want to check this out, download a clean CP/M 2.2 from the SIMH site, copy just one large (eg >16k) file to the "i" dsk, then examine the i.dsk file using a hex editor. There is something really satisfying about cracking a code like this. What are the chances of us mere mortals ever really understanding fat32 for instance?
If you can crack the "fat" code, you can read and write directly to a SD card file. And that means that Wordstar can read and write a .txt file straight off an SD card. That gives you multiple free text editors for the Propeller chip.