No scoffing from this direction.

I'm somewhat surprised with the Vce(sat) for the opto, but if that's what you've measured then that's what it is. Personally at that stage I'd be using change-over contacts for single-switched inputs and using NC=gnd, NO=5V just to yank that signal line to one rail or the other (nothing left floating).

For the nay-sayers, what Kitwn's post is providing is belted-and-braced robust and deterministic behaviour. There's too many problems in the world to save pennies on uncertainty on our machines.

Perhaps I'm feeling humbled after a day's worth of trying to get my control box switching automatically between two machines... and making some stupid assumptions... right, where's that damned scope gone...