Curious, what's at the far end? How fast is this supposed to go?
Guessing it must be pretty slow (kbps tops?), given the line length, weak drive strength, and amount of filtering required to get a reasonable logic signal out of it. I hope the receiver is high impedance.
If you like belt-and-suspenders approaches, I'm fond of the complementary bipolar inverter. It is what it sounds like: make a CMOS inverter, two transistors; but sneak in BJTs instead, and the requisite biasing (base dividers) of course. In this case, one last step: emitter resistors, so the current is limited.
The output drive is rail-to-rail, with an on-resistance a little higher than the emitter resistance, until the current limit is reached.
Speed is way more than needed, and can be ~10 ns with faster transistors and speed-up caps in the base dividers. Slow is probably the better way here, and can be done by just loading the output with a cap to ground (since, it's current limited, no problem eh?). Follow that with a nice sized ferrite bead (say 330 ohm, >= 0603) and you've got TX side filtering handled.
The collectors have a lot of compliance voltage up and down (no body diode to get snagged on), so you can use whatever surge protection is appropriate. Say, a single SMAJ5.0A, more than enough. At peak surge voltage (10-20V), there's a fair amount of stress on the transistors, but everything's hanging by resistors, so it rides through no problem (just don't expect it to survive very long if cross-wired to, say, +/-12V or more).
What about ground loop, is that any concern..? Is there isolation? Could ground loop offset or surge be a problem? That's more of a RX problem but at a high enough level (e.g. induced lightning surge), it can even be a TX problem.
Ed: also, is this a one-off, or are there, like, hundreds or thousands of these links? If scale or size matters, a discrete approach will be annoying, and an IC preferred. I'm not sure offhand what kinds of device may serve the same purpose (*maybe* but probably not an LVDS transmitter? LED driver?) but it would be interesting to hunt around.
As you note, RS-422 transmitters are basically full logic drivers, they're made to dump lots of current, and use large transistors to handle the relatively high power dissipation. (And yes, leaving off the opposite phase pin is okay.) Would be close enough to use that with a series resistor, really; 74HC is close to 40 ohms by itself, so you can still improve on that while keeping the driver happy by using an external resistor of comparable value.
Tim