Are you familiar with the CMOS latchup effect, where the device starts smoking from induced SCR trigger effect from stray electro-static discharge (ESD) from someone who grabs a cable end after walking on an office carpet with plastic soled shoes? I have seen this too many times to know that the built-in diode clamp protection is not enough. Why? because ESR rise times are on the order of 10~50 ps (pico Sec) and the diode clamp rise time is ~ 1nS and if the signal line exceeds the supply rail by more than 0.5V you are pooched.
That's why it's best to stick with bipolar robust designed RS232 drivers.
BTW I designed a split PAX unit as a contract for a Canadian telecom data company in 1983 one week, but now a days, you got all these CPLD's and bigger devices that can do the the MUX/ DEMUX much easier. I dont know if they ever pursued my conceptual design, but it earned our company 50K$ for the work.
MOSFETS will work as long as your TVS clamp/filters dont slow down slew rate with an ultrafast recovery diode (Expensive). So I suggest standard RS232 or choose the high amplitude Limited distance Data set drivers for RS232 to go a few Km. But really it makes more sense to use a VPN over ethernet at 10Mbps rates and then MUX-DEMUX to hex RS232 for local traffic using Xilinx etc for the logic/ sync and use ethernet or CRC.
How many ports and what rates? and what distance? Autobaud detect? Flow control by Xon Xoff or RTS CTS DSR DTR? or what? need specs please.
I guess it's safe to say,.. "the rapid rise of TCP/IP relegated many of Gandalf's products to niche status, and the company eventually went bankrupt in 1997."(Gandalf were famous for LDS data sets/modems etc.) You might find some kicking around in surplus.