We've already established that the 8-port MCU won't fit in the application board.
I have no such knowledge. If an 8 port MCU will not fit where is the room for the 'expansion' chips?

A small, cheap, slave MCU? Well that's a pregnant thought!
the birth has happened. My current major device has a "A small, cheap, slave" to process the keyboard "A small, cheap, Cypress chip" as a USB master "a not so small not so cheap F13x" as a communicator in addition to the f12x processor.

Parallel latches, well, they'll obviously work if one uses an external memory bus. You've pointed out, however, that you don't like discarding two ports, so that only applies if the external bus is already present.
Of course, that should be self-evident. The issue here is that it will be extremely rare that you need parallel expansion unless you have already applied external memory.

CPLD is quite a flexible solution and will work. I doubt anyone will want to use an ASIC, since the they routinely cost on the order of $2E6US for the first one.
I have worked with ONE such, but I should have said CPLD/FPGA

The "WSI" chip you keep mentioning. You mean the ST (u)PSD813-series? Which one? Please be specific. Remember, WSI doesn't exist any longer. What do they cost? I can't find any of them for sale anywhere.

FOR THE LAST TIME I included the PSD chips (which I have used in the distant past) for the sake of making a complete list not to be dragged into performing a wed search.

Jan, can you (who evidently use the flash basewd psd chips) answer these questions, just to shut Richard up,

Erik