From my point of view, the issue is just as Gyro described:
I would take the +/- 50V with a pinch of salt. It's really an intermittent input protection voltage rather than a usage one.
Those Salaea 16 clones (well semi-clones) do include separate ESD protection diode packages (unlike the 8 bit ones), just as well as the inputs go into an FPGA. I think the input series resistor values will be too low for countinuous operation at RS-232 levels without causing excessive loading on your signals and high internal dissipation on the input circuits.
I can't tell you how those are designed but I'm willing to bet that those devices simply have internal clamping diodes that start conducting when exceeding a small voltage and series resistors to limit the loading on the data lines you want to measure. They might work but there's no guarantee, especially if the series resistor is too small. Example:
http://sigrok.org/wiki/Mcupro_Logic16_clone - the series resistor is only 510 ohms. At +/-12V, that'll draw ~20mA on the data lines, potentially altering the signals.
If you want, you could get one, try it out and if it doesn't work, add a MAX232 or zener+resistor "frontend".
Personally, I'd just get
this board and make a little daughter board to stack on top (perfboard is perfect), featuring a bunch of 4V7 zeners and 10k resistors. If you want to have a ready-to-use solution, it'll either be a scanaplus or trial and error.
Edit:
http://sigrok.org/wiki/File:Xl_logic16_100m_black_pcb_top.jpg shows yet another variant and there you can see 472 resistor arrays, so there are 4k7 series resistors present, which would actually work for you. However, it doesn't (yet?) have sigrok support. Looks like a gamble to me...