So here's a teaser.
3 reed switches are mounted on the track; the yellow stuff is modeling clay to temporarily hold items in place. On the truck is a holder for 3 Nd tiny disc magnets. So either 0, 1, 2, or 3 magnets can be installed with idea being this allows identification of up to 7 different devices since there's effectively a 3-bit digital address (000, 001, 010, 011, 100, 101, 110, 111) with 000 being the idle case. In this case the Nd magnets are 3mm diameter x 1.5mm thick - a few pennies a piece.
I had these reed switches in my parts stash from DigiKey but I see they are obsolete. Something like these 50 cent switches are probably equivalent based on a cursory look at the datasheets.
Here's a proof-of-concept:
The video first shows how there's sufficient position "resolution" such that a magnet can trip one-and-only-one reed switch without tripping the adjacent reed switch. Obviously this is a necessary condition for the scheme to work. Of course all this is possible with O-gauge dimension because of the ridiculously tiny size of rare-earth, high-power Nd Neodymium magnets.
Without going thru each of the 7 combinations (3-bit addresses), I just show the "101" address to trigger the two outer sensors but NOT triggering the center sensor. This of course is another necessary condition.
Notice how the LED indicators remain on for a second or so after the train zooms by the sensors. This gets back to my previous comments about aperture time. Reed switches can respond incredible quickly (less than a thousandth of a second!) but if the electronics is not looking during that time, it will miss the event. With the addition of a 5 cent capacitor, the triggers can be extended to give whoever is looking more time to get it together.
Again, this is just a proof-of-concept slapped together on a rainy afternoon. There are many i's to dot and t's to cross. I was just curious if this would even pass the laugh-test, release magic-smoke, blow up in my face, whatever. If taking the next step I would change to Hall sensor chips instead of reed switches (smaller, solid-state no-moving-parts, no "fragile" glass housings, etc.).
As already discussed, this method is apples-and-oranges with the NFC method. OTOH, the engine-side equivalent to the RF tag is less than 10 cents (for up to 3 magnets). The track-side "interrogator" is less than $2 consisting of 3 magnetic sensors (reed switch or Hall sensor IC chip) and a handful of 5 cent resistors and capacitors.
