What would people's opinion on using a 5V FPGA (or possibly large-I/O-count CPLD) for a new, but limited run in-house-product in this day and age be? This would be a USB-tethered lab/ATE/field-service/etc "universal interface" intended to replace some aging parallel-port driven devices not supported by modern PCs (especially laptops).
The reason for a 5V part is that performance demands are very low, but it needs to interface with up to about 100 bits of 5v instrumentation control bus, driving as well as receiving. If I use a part with a 3.3v I/O voltage I need level translators. That means a much more complicated board. I realize not using translators exposes the FPGA to greater risk of damage from the outside world, but now that we have a good rework station changing out one blown FPGA isn't that much worse than changing out multiple blown translators.
I'd also like to avoid the level translators to ease reconfiguration - I believe I can use the same FTDI USB chip to JTAG program the FPGA in addition to talking to it, and that would let my device driver download an appropriate FPGA configuration for any of our applications, which might need to change which pins are inputs or outputs on a bit-by-bit basis rather than the byte-by-byte basis most translators would support.
Other than making sure to buy enough chips + spares up front, any reason not to use an older part like an original spartan? Any particular 5v devices with 100+ I/O's in non-BGA packages that might be more likely than others to hang around?