This instance was for a power input rather than ,say, a switch input. If you look at the inputs to industrial PLCs you have an opto coupler with a zener and a resistor. If the zener fails the resistor takes the brunt of it - the resistor is 1k or thereabouts so it wont get too hot - unless someone puts 240vac into it!

For industrial equipment I try to have the devices fail passively - if you wire the DC volts backwards (very likely) then there shouldn't be smoke - either a fuse blows or the equipment tolerates it - thus my use of a polyswitch. On one piece of equipment I have 24vdc along with RS485 comms - it was suggested to me 'what happen if the connector is wired backwards'? I modified my design to tolerate this.


Back to the issue of two boards - if you can fit everything onto one board, you save on testing (1 test rig instead of 2), one board to assemble, one board to handle, one board to install, one board to repair, one parts list, one spare part........... I find you have to look beyond just the item cost of the pcb. The issue of labour (especially in testing) tends to outweigh the pcb cost. Every situation is different but I find engineers tend to have a myopic view of cost - sometimes cutting cost in one area increases the cost in others - it becomes a bit of a ying/yang sorta thing.