In my industry, we mostly use the term to refer to mechanical features, typically keyways, that prevent connectors from being mated incorrectly.
Suppose you have a module with 50 pins worth of connectors, but because some of the signals are in different harnesses which get installed at different times, you can't just use a 50-pin connector. For this example let's say it's sensible to break it up as 20+20+10.
You wouldn't use two identical 20-pin connectors since they carry different signals. You probably do want to use the same family of connectors so they use common pins and have a common board footprint. So you get connectors which are the same except for having different keyways, and are often molded of different colors of plastic, by convention.
If you've ever been working on a vehicle and seen identical-looking connectors where one's black and one's gray, look closer. Along with a color difference, there's a notch on the housing in a different place.
This increases the number of parts that must be stocked, but the decrease in assembly errors is worth it. (And they all share the same tooling, so the manufacturing complexity isn't bad.)
Note that this isn't required if the connectors aren't candidates for mismating in the first place. If they appear in completely different places on the harness, then it's totally fine to reuse the very same connector for the amplifier speaker signals in the trunk, and the steering column (turnsignal stalks and stuff) module signals up front. This reduces parts count without increasing errors.
When I was designing boards, we had 1-2 standard types of connector we were allowed use, that had an even number of pins. So we would use a different number of pins for each connector on the board, even if some of the pins were not used. It looked kind of silly to see 4 wires in a 12 pin connector but it was better then getting them mixed up.
Yup, I do this myself just to reduce the number of parts I stock.
I'll try to plan uses for all the pins in advance, like a deluxe version, and then cut down what's needed for the thing I'm doing now. Often I can actually add something useful (like an extra copy of a signal to reduce the need to splice in the harness) at no cost, just by thinking ahead a little.
Another thing I'm proud of recently, is having an uncommitted relay on a board, whose function could then be defined by how the harness was wired. Rather than try to anticipate how it would be used, I just figured it'd be handy. Provided an extra ground too, so you could just put a hairpin wire in the connector to ground one side of the relay since that's a pretty common usage model. And for builds where we knew we wouldn't use it, just DNP at assembly time.
The other solution is to make it trivial to both fix and detect errors, by using the same pinout so nothing gets damaged, and letting the computer tell you if everything is right.
But with a car that's not really practical as things may not be accessible and easy to swap around.
Nor powered up at all stages of assembly.
I didn't spend a ton of time at the plant, but there was one notable visit where I got to camp out at the station where, for the first time, a 12-volt power supply alongside the assembly line would be temporarily patched into the half-completed vehicle. This allowed all the modules that'd been added so far, to boot up and do some self-checks. There were certain missing sensors and communication errors that were expected at that point and would be ignored, but plenty of errors that _would_ be raised so they could be fixed while they were still easy-ish to reach, before the seats and trim panels started going in.
Worse than different signals, sometimes they carry power rails in different locations, so plugging the wrong connector in can sometimes destroy electronics or start a fire.